Edit: Fixed toBitSet() ignoring its argument and returning every track in the Edit

[drowaudio]

Fixes #399

Problem

toBitSet (const juce::Array<Track*>&) (modules/tracktion_engine/model/edit/tracktion_EditUtilities.cpp:250) used its argument only to reach the Edit, then looped over getAllTracks() instead of tracks:

for (auto t : allTracks)          // <-- should be `tracks`
    if (int index = allTracks.indexOf (t); index >= 0)
        bitset.setBit (index);

The result therefore never depended on what was asked for - every track in the Edit got a bit. The mask is live: Renderer::Parameters::tracksToDo becomes CreateNodeParams::allowedTracks (Renderer.cpp:38), so a caller asking for a subset silently rendered or measured the whole Edit.

It stayed hidden in-engine because Renderer::renderToFile() passes getAllTracks (edit) (for which "every track" is the right answer), and the one in-engine call passing a subset (tracktion_NodeRendering.test.cpp:203) is on an Edit with a single audio track. The engine's own tests had already worked around it: tracktion_EditNodeBuilder.test.cpp and tracktion_Plugins.test.cpp each carried a private getTracksMask() built from Track::getIndexInEditTrackList() rather than calling the shared helper.

Plan

1. Fix the loop - iterate tracks rather than allTracks in toBitSet(). allTracks.indexOf() still supplies the index, so the bit numbering is unchanged and tracks from another Edit are skipped.
2. Tighten the doc comment on tracktion_EditUtilities.h to state what the bits mean, where the Edit comes from, and that foreign tracks are ignored - the vague wording is part of why the bug survived.
3. Add tracktion_EditUtilities.test.cpp (new ENGINE_UNIT_TESTS_EDIT_UTILITIES flag in tracktion_TestConfig.h, included from tracktion_engine_model_1.cpp) covering: a single track, an arbitrary subset, all tracks, an empty array, tracks belonging to another Edit, agreement between the bit index and Track::getIndexInEditTrackList(), and a toTrackArray() round-trip.
4. Delete the two getTracksMask() copies in tracktion_EditNodeBuilder.test.cpp and tracktion_Plugins.test.cpp and call toBitSet() instead - those helpers exist only to route around this bug, and keeping them would leave the fix untested by the suites that actually render subsets.
5. Record the behaviour change in BREAKING-CHANGES.md: host code that passed a subset now gets a subset, which changes what renders.

Verification

• The new test fails against the old loop (15 failing assertions) and passes with the fix, so it is not vacuous.
• TestRunner Debug builds clean, and the suites touching the changed helpers pass: tracktion_EditUtilities.test.cpp, tracktion_EditNodeBuilder.test.cpp, tracktion_Plugins.test.cpp and tracktion_NodeRendering.test.cpp - 12 test cases, 543 assertions, 0 failures.
• Nothing else in-engine feeds toBitSet() into a bitset: TransportControl.cpp:969, AudioTrack.cpp:1094 and RenderOptions::getTrackIndexes() all build their masks directly, and Renderer::renderToFile() passes every track, so its result is unchanged.

Note on behaviour

test_utilities::expectPeak() in tracktion_TestUtilities.h has no in-engine callers but is part of the public test utilities - host tests using it against a subset of tracks will now measure that subset, which is the documented intent but may shift existing expected values.