HamiltonDeck: correct track count & rename rail->track

[BioCam]

The STAR v1 integration addresses carriers by track throughout. This lands two corrections to how PyLabRobot describes a Hamilton deck on their own first, so that PR stays smaller.

The problem

PyLabRobot models a STARlet deck with 32 positions and a STAR with 56, but the decks have 30 and 54 tracks. The two extra positions are where the waste block stands, so rails=31 and rails=32 were accepted although no carrier can stand there.

The positions are also misnamed. Hamilton calls them tracks: the lanes between the slideblocks numbered left to right across the deck, 22.5 mm apart. PyLabRobot calls them rails.

PR Content/Solution

• STARLET_NUM_TRACKS = 30 and STAR_NUM_TRACKS = 54, and track= on assign_child_resource is bounded by them.
• num_tracks, track_to_location and track_for_x_coordinate replace num_rails, rails_to_location and rails_for_x_coordinate. The old names still work and raise DeprecationWarning; on a STAR deck num_rails and rails= keep the counts and bounds they had.
• num_tracks is the first argument, where num_rails was. A subclass may implement either track_to_location or rails_to_location, and passing both num_tracks and num_rails raises ValueError.
• The docs and the legacy STARBackend move to the new names.

Behaviour: every position is unchanged - track n is at the x rail n was - and decks saved on main load with every resource where it was. A deck now serializes num_tracks, which older versions cannot read, and a count passed positionally to HamiltonSTARDeck is read as tracks.

Tests: nine added in hamilton_deck_tests.py (the new bounds, each deprecated name, a deck saved with num_rails, a subclass implementing only rails_to_location, both counts refused); the 48 existing call sites move to track= with no expected value changed. ruff format, ruff check --select I, ruff check, typos and mypy pylabrobot --check-untyped-defs are clean; the full suite passes (3261 passed, 4 skipped, 802 subtests).

🤖 Generated with Claude Code