PI stage: connected only after a full initialize, a reset id, a wait for the controller, and fake-GCS2 tests

[kalidke]

The PI stage (PIStage, C-867) fixes from the #69 review, as one PR into main. They are for 0.2.6 and were kept out of 0.2.5.

This branch carries #72 (the 0.2.6-DEV bump) merged in, because a PR into a released main must otherwise raise the version. Until #72 merges, its one commit shows here too.

Fixed

• shutdown could close another object's connection. id defaulted to 0, a valid GCS id, and was never reset. It now defaults to -1, and shutdown resets it; a failed close keeps the id so a later shutdown can retry.
• initialize reported the stage connected before it was (review F2). connectionstatus was set before the connect, and a failed close after a failed reference left it true, so a retry answered "already initialized". It is now set only after the whole sequence succeeds, and every step after the connect is inside the cleanup, which closes the connection and rethrows.
• The stage panels' initialize buttons (1D, 2D and 3D) log a failed initialize and return, instead of throwing out of the Makie click callback (review F4). They also skip the position read when initialize returned without connecting.

Changed

• initialize waits for PI_IsControllerReady after the reference move, before polling PI_qFRF, as PI's samples do (review F3). Both waits share REFERENCE_TIMEOUT_S (60 s). Not yet run on hardware; this needs a rig check on the C-867.
• Docstrings: initialize says what it does; referencemove is the PI_FRF start. The stale TODOs are gone, and so is the "1 mm/s" comment (review F6).

Tests (F7)

• A seam: pi_stage/gcs2.jl wraps each of the 19 GCS2 calls PIStage makes as a function, so test/pi_stage_fake_sdk.jl (FakePIStage) replaces every one. Calls the tests don't model (PI_MOV, PI_HLT, PI_STP) throw, so no test can reach a DLL. No call's types changed.
• test/pi_stage.jl has 8 cases:
  • the default id, and shutdown before initialize;
  • the success order FRF, then IsControllerReady, then qFRF, with connectionstatus still false during the reference;
  • a refused FRF;
  • a controller that is never ready;
  • an axis that is never referenced;
  • a throw after the reference;
  • a failed close, then a retry;
  • shutdown's id reset.
• Local suite at this head (kitt, Julia 1.13): 1841 pass, 3 broken, 0 fail. The lab/tests record follows as a commit status.

Not yet run on hardware. MicroscopeAdapt's rig builds PIStage().

🤖 Generated with Claude Code

[kalidke]

This ships in #76 (MicroscopeControl 0.2.6), the single 0.2.6 release PR, as a merge commit of this PR's accepted head. Please leave this PR open: GitHub marks it merged when #76 lands.