Clamp ally-team widget index in the map editor's teams editor

[kylelutze]

Summary

• The map editor's teams editor converted the header's 1-based ally-team number to a 0-based widget index by bare subtraction. A header with 0, or a value above the map's team count, gave an out-of-range index; MultiTextButton::setIndex uses vector::at, so the dialog threw and crashed on open.
• Promotes the clamp helper already used by the custom-game options screen into a shared header-only AllyTeamWidgetIndex.h and uses it at both sites. Well-formed values still map to value - 1; only headers that previously threw now display ally team 1. Nothing written back changes.
• Names the teams editor's widget return-code bases (PLAYER_ACTIVE_BASE, COLOR_BASE, AI_SELECTOR_BASE, ALLY_TEAM_BASE) instead of hand-typed 100/200/300/400, and documents generateGameHeader's inverse mapping.

Verification

• scons -j16 clean.
• tests/ally_team_widget_index_test.cpp (standalone, compile line in its header) passes all cases: 0, above-range, and 255 clamp to 0; 1..teamCount map to 0..teamCount-1.
• Workflow 1 replay check: cpp-refactor.replay byte-identical (editor UI only, no sim path touched).

Manual check

Editor → open teams editor. Before: a map whose game header carries an ally-team number outside [1, numberOfTeams] crashes on dialog open. After: that row shows ally team 1. Well-formed maps look identical.

[Giszmo]

Reviewed at head 4214b84, tested merged onto master bfc8fd7. Approve, no blockers.

Verified

• The crash is real: MultiTextButton::setIndex uses texts.at() and the ally widget only has numberOfTeams entries, so an ally number of 0 or above the team count threw on dialog open. The clamp is the same helper already used in CustomGameOtherOptions (4f13a60); well-formed values and the index + 1 write-back in generateGameHeader are unchanged.
• The named return-code constants keep exactly the old 100/200/300/400 ranges; onAction logic is untouched.
• scons -j16 on master + PR builds clean with no new warnings in the touched files. tests/ally_team_widget_index_test.cpp compiles and passes with the command in its header. The branch is a few commits behind master but merges clean.

Non-blocking notes

1. Reachability: the editor's own add/remove team already rebuilds the header from scratch (MapEdit::regenerateGameHeader, src/map/edit/MapEditClicks.cpp:391), so this only fires for a map or game file whose stored header is already out of range. Fine as hardening; the PR text does not overclaim.
2. Fallback choice: every out-of-range value collapses to widget 0 = ally team 1, so after OK those teams end up allied with team 0. Falling back to the team's own index (the reset() no-alliance layout, always < teamCount for a valid team) would be a more neutral default. Keeping 0 is defensible for consistency with the CustomGameOtherOptions site.
3. Test placement: the repo convention is test/ (singular) with CppUnit fixtures built by test/SConstruct into TestsRunner, which CI runs. This file lives in tests/ and is hand-run only, so CI never executes it. Suggest moving it into test/ as a fixture (FilenameStripTest.cpp is a minimal template).
4. Pre-existing, out of scope: the same malformed file also feeds color[i]->setSelectedColor(teamNumber), which ColorButton stores unchecked, so the dialog now opens but OK writes that team number back.

[kylelutze]

Thanks for the review. Note 3 addressed in 0b0f598: the test is now a CppUnit fixture in test/AllyTeamWidgetIndexTest.cpp, registered in TestsRunner (158 tests pass locally). Note 4 tracked as #188. Keeping the index-0 fallback (note 2) for consistency with CustomGameOtherOptions.