Validate WorkerClient batch coordinates before image access

[arjunrajlab]

Summary

• validate Batch XY/Z/Time coordinates against dataset IndexRange before image access
• report that batch positions start at 1 and include the dataset's valid range
• report malformed batch syntax through sendError instead of leaking parser exceptions
• validate stacked channel selections and preserve parsed batch ranges across repeated preflight checks
• run the Cellpose-SAM preflight before loading the GPU model
• document Cellpose-SAM's 1-indexed batch contract and record the scoped Compose test-runner gap discovered during verification

Root cause

Batch fields are entered as 1-indexed positions and converted to zero-based coordinates. The production value Batch Z: "0-34" therefore became [-1, ..., 33]. WorkerClient passed -1 directly to coordinatesToFrameIndex, where the dataset map raised KeyError: -1 without first sending a useful worker error.

The earlier Failed to get docker network hostname-resolution message was non-fatal: the container started, detected CUDA, and reached image loading before the coordinate failure.

Impact

Invalid batch coordinates now fail before frame lookup, with a message such as: Batch Z contains invalid position 0. Batch positions start at 1; ... valid range is 1-35. Cellpose-SAM performs this validation before constructing the large model, and all workers using WorkerClient.process() inherit the shared guard.

Validation

• pytest -q worker_client/tests — 19 passed
• pytest -q annotation_utilities/tests — 61 passed
• pytest -q workers/annotations/cellposesam/tests — 13 passed
• python -m py_compile for the changed Python modules
• git diff --check

./build_workers.sh --build-and-run-tests cellposesam successfully rebuilt the shared base images but did not run a worker test because cellposesam/cellposesam_test are absent from Compose; the script nevertheless returned zero. That pre-existing build-runner issue is documented as TODO-005 in this PR.