configs: remove duplicate INI keys (first occurrence wins)

[PeterStolz]

What

Several shipped sim/demo configs define the same INI key more than once within one section. LinuxCNC only reads the first occurrence, so the later lines are silently ignored — in two cases hiding the author's intent. Split into two commits:

1. conflicting duplicates (different values — a value was silently dead)
2. redundant duplicates (same value — pure no-op de-clutter)

How I found it

I'm building a LinuxCNC VS Code extension (HAL/INI/G-code language support) — https://marketplace.visualstudio.com/items?itemName=PeterStolz.linuxcnc-lsp — and its INI linter flagged duplicate keys across a number of in-tree configs, so I went through each to confirm they're real.

Why the second value is ignored

IniFile::findTag(tag, section, num) defaults to num=1 and returns the first match in file order (src/emc/ini/inifile.cc); there is no concatenation for scalar keys. The same is true of the inivar CLI used by scripts/linuxcnc and of linuxcnc.ini(...).find() used by the GUIs. Confirmed on a 2.10 run-in-place build:

$ inivar -ini configs/sim/axis/histogram_demo.ini -var MAX_LINEAR_VELOCITY -sec TRAJ
1.2          # first wins; the file's later 1.2345 is ignored

>>> linuxcnc.ini(".../woodpecker_xyzab.ini").find("DISPLAY","GEOMETRY")
'XYZABCUVW'  # findall() -> ['XYZABCUVW','xyzab']; the intended 'xyzab' was dead

Commit 1 — conflicting duplicates

Behaviour-changing (restore intent):

• sim/axis/gladevcp/probe.ini — [RS274NGC] SUBROUTINE_PATH was set twice, so the second value (../../nc_files/gladevcp_lib, the gladevcp demo's O-word subs, per its own comment) was never on the path. Merged into one colon-separated SUBROUTINE_PATH.
• sim/woodpecker/woodpecker_xyzab.ini — [DISPLAY] GEOMETRY = xyzab (added for this XYZAB variant) was overridden by a leftover XYZABCUVW; made xyzab effective.

No-op (the removed value was already the ignored one):

• woodpecker_xyzab.ini — duplicate [RS274NGC] PARAMETER_FILE (metric_parameters.txt dead; keeps woodpecker.var).
• woodpecker_xyza.ini — a pasted-twice [DISPLAY] spindle-override block.
• histogram_demo.ini, ini_hal_demo.ini — stray second [TRAJ] MAX_LINEAR_VELOCITY = 1.2345.
• sim/axis/vismach/5axis/bridgemill/5axis.ini — redundant [TRAJ] MAX_ANGULAR_VELOCITY = 360 (identical to 360.0).

Commit 2 — redundant (same-value) duplicates

Pure de-clutter, no behaviour change:

• [TRAJ] MAX_LINEAR_VELOCITY = 58 in sim/axis/gantry/gantry.ini, gantry_jjog.ini, sim/qtaxis/gantry/qt_gantry.ini, sim/qtvcp_screens/qt_gantry.ini, qtdefault_gantry.ini.
• [RS274NGC] HAL_PIN_VARS = 1 in sim/axis/vismach/scara/scara.ini, sim/qtaxis/non-trivial/scara/scara.ini, sim/qtvcp_screens/non-trivial/scara/scara.ini.
• [DISPLAY] GEOMETRY = XYZCBW in sim/axis/vismach/5axis/bridgemill/5axis.ini.

Not included

The Sherline Sherline4Axis_{inch,mm}.ini configs have a # for gui only block where the first [TRAJ] MAX_LINEAR_VELOCITY (25) wins over a more realistic second value (.36 in/s · 8 mm/s). Fixing that changes the effective max velocity — a judgement call — so it's left for separate discussion.