fix: apply channel z_offset to stage position z when there is no z_plan

[hinderling]

Problem

Channel.z_offset is only applied in _xyzpos() when z comes from a z_plan. When z comes only from the stage position, the elif position: branch uses position.z and drops the offset. So the same sequence behaves differently depending on whether a z-stack is enabled. User-visible: pymmcore-widgets' channel table has a Z offset column, which has no effect unless a z-plan is also set.

import useq

seq = useq.MDASequence(
    stage_positions=[(0.0, 0.0, 0.0)],
    channels=[{"config": "a"}, {"config": "b", "z_offset": 5.0}],
)
[e.z_pos for e in seq]  # [0.0, 0.0] on main, expected [0.0, 5.0]

With z_plan={"range": 2.0, "step": 1.0} the offset is applied (channel b gets 4/5/6 vs a's -1/0/1). Channel.z_offset is documented as a relative Z offset in microns, with no mention of requiring a z-stack. pymmcore-plus doesn't read z_offset itself (the offset is meant to be baked into event.z_pos here), so nothing else applies it.

Fix

Add the channel z_offset in the elif position: branch too. The two branches are mutually exclusive, so the offset is never applied twice. If the position has no z (position.z is None), z_pos stays None: there is nothing to offset.

Test

Added to tests/test_sequence.py:

• test_channel_z_offset_without_z_plan: fails on main ([10.0, 10.0] != [10.0, 15.0]), passes with the fix.
• test_channel_z_offset_with_z_plan: offset applied once (unchanged behaviour).
• test_channel_z_offset_without_position_z: z_pos stays None.

Behaviour change

Sequences with a nonzero z_offset and no z_plan will now move Z per channel, where before the offset was ignored. z_offset defaults to 0.0, so only sequences that set it explicitly are affected.

Not covered (possible follow-ups)

• No absolute z to offset. With no stage_positions, or a position without z, z_pos stays None ("don't move Z") and the offset is still dropped. The docstring describes the offset as relative to the current position, which an event can't express without a notion of the current z (there is already a TODO about this in _xyzpos). Supporting it would likely mean resolving the offset in the engine.
• Position sub-sequences. For stage_positions=[{"x": 0, "y": 0, "z": 3, "sequence": {"channels": [{"config": "b", "z_offset": 5.0}]}}], z_pos is still 3.0: _position_offsets forces z to the parent event's z whenever the sub-sequence has no z_plan, which overwrites the sub-sequence channel's offset. Separate code path, so I left it out of this PR.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.63%. Comparing base (96299a2) to head (7e39049).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #274   +/-   ##
=======================================
  Coverage   93.62%   93.63%           
=======================================
  Files          33       33           
  Lines        2589     2591    +2     
=======================================
+ Hits         2424     2426    +2     
  Misses        165      165           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[tlambert03]

thanks!