fix(instagram,facebook): prevent broken page-grant channels and guide reconnection

[giladresisi]

Problem

Connecting an Instagram (Facebook Business) or Facebook Page channel while granting the wrong Facebook page silently stores a broken token — undefined___<user-token> for Instagram — and every publish then fails with Graph code 190 disguised as a misleading "missing permissions" error. There is no reconnect prompt and no way out short of support.

This is easy to hit for two reasons:

1. The account picker also offered pages discovered via the Business Manager API that were never granted to the app in the OAuth dialog.
2. Facebook's granular consent lets users grant an Instagram account without its linked page (they're independent asset lists).

Reproduction

Connect an Instagram (Facebook Business) channel and, in the Facebook dialog, grant a page that is not the one linked to the Instagram account. Postiz still offered the IG account in the picker; selecting it connected a channel whose token is undefined___…; every post then failed with "The account is missing some permissions…" — with no indication the channel needed reconnecting.

Fix

Prevention (silent, no new UI in the picker):

• pages() (Instagram) now verifies every candidate page with the same page-node access_token call the save path uses, and only offers pages the app can actually publish through. (The Business-Manager edge can report a token the page node refuses, so the node call is the only check that keeps the picker and the save consistent.) The Facebook provider filters on its edge tokens, which its own save path reads.
• auth_type=rerequest on the OAuth URL, so declined permissions/assets are re-prompted and a bad grant can actually be repaired on reconnect.

Recovery for already-broken channels:

• Graph code 190 is now classified as refresh-token: the next post from a broken channel marks it refreshNeeded and emails the user; subsequent posts keep failing with a reconnect email until fixed.
• The refresh-error email now explains that no posts can be published until the channel is reconnected, and that re-adding the same account replaces the broken connection and keeps all scheduled posts (no deletion needed).
• updateIntegration() clears refreshNeeded when a channel is re-added over an existing one, so "Add Channel with the same account" fully heals a flagged channel.

Guidance UI (Instagram/Facebook only):

• A "Before you continue" popup before every redirect to Facebook — on Add Channel, on reconnect, and on half-connected channels — explaining to select the Facebook page the Instagram account is connected to, and that changing a previous selection goes through Facebook's "Edit settings".
• Half-connected IG/FB channels restart the OAuth flow when clicked (the in-app picker is a dead end when the grant is wrong); other providers keep the in-app continue flow.
• "Back to Postiz" button on the full-page empty selection state (previously a dead end with no navigation).
• Connect errors are surfaced as a toast in the page-selection modal instead of silently closing it.
• New i18n keys translated across all 16 locales.

Shared-grant hazard (second commit)

Facebook page grants are per Meta app and per user, not per channel: re-authorizing any Instagram/Facebook channel with a narrowed page selection revokes the unchecked pages' grants and invalidates their page tokens — silently breaking every other Postiz channel that depends on them. Verified experimentally in both directions:

• Reconnecting Instagram while granting only the Instagram-linked page → the existing Facebook page channel started failing with an unmapped code-190 error (surfaced as "Unknown Error").
• Reconnecting the Facebook page while granting only that page → the Instagram channel started failing with the code-190 permissions error.
• Re-granting both pages fixed both channels, with no changes to the channels themselves.

The code-190 → refresh-token classification in this PR already routes such casualties into the reconnect flow at their next post. The second commit additionally adds a line to the "Before you continue" popup warning users that their page selection applies to all their Meta channels and that unchecking pages used by other channels will break them (translated across all locales).

Post-fix validation (manual, real Facebook/Instagram accounts)

• With the wrong page granted: the IG account is no longer offered — empty selection state with a working "Back to Postiz" button.
• Posting from a previously-broken channel: post fails once, the channel is flagged (badge + "Reconnect channel" menu), and a single clear email is sent.
• Reconnecting via the popup → Facebook screens → correct page: the channel heals in place, refreshNeeded clears, and all scheduled posts survive.

🤖 Generated with Claude Code

[postiz-contribution]

Contribution-checker quality warning
Heuristic score: 0/100 (low). This is a non-blocking warning surfaced by the project's quality settings.

Heuristics that flagged:

• Wall-of-text PR body: 3680 chars (>2500)
• Excessive inline code references: 10 inline refs (>3)
• PR doesn't use the repo's PR template: 5 required checkbox items missing (max 1, match ≥80%)
• Body adds too many extra headers beyond the template: 4 extra headers (>1)
• AI watermark phrase: Matched: "Generated with Claude Code"
• Commit message too long: Longest: 2866 chars
• Excessive added comments: 30 added comment lines (ratio 0.09)
• Missing final newline

If this is a genuine contribution, please add detail to your PR description and tighten the diff scope before reviewers look at it.

[postiz-agent]

✅ Snyk checks have passed. No issues have been found so far.

Status	Scan Engine	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues
✅	Licenses	0	0	0	0	0 issues
✅	Code Security	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.

[postiz-contribution]

Contribution-checker quality warning
Heuristic score: 0/100 (low). This is a non-blocking warning surfaced by the project's quality settings.

Heuristics that flagged:

• Wall-of-text PR body: 3680 chars (>2500)
• Excessive inline code references: 10 inline refs (>3)
• PR doesn't use the repo's PR template: 5 required checkbox items missing (max 1, match ≥80%)
• Body adds too many extra headers beyond the template: 4 extra headers (>1)
• AI watermark phrase: Matched: "Generated with Claude Code"
• Commit message too long: Longest: 2866 chars
• Excessive added comments: 30 added comment lines (ratio 0.08)
• Missing final newline

If this is a genuine contribution, please add detail to your PR description and tighten the diff scope before reviewers look at it.