feat(lif): extract per-scene acquisition/instrument block (date, microscope, imaging_method) into audit + inspect

[ferrinm]

What to build

Parse per-scene acquisition + instrument metadata from LIF scene XML and surface it in per_scene[i].acquisition (or the block name locked in ADR-0007) in both the convert audit and inspect() output.

Fields to extract

Per scene, mapping to Aperture BigQuery columns:

• acquisition_date — timestamp of scene acquisition. LIF stores this in Attachment[TimeStampList] / TimeStamp elements on the scene.
• microscope_model — brand + model of the instrument, e.g. "Stellaris", "SP8". LIF HardwareSetting / ATLConfocalSettingDefinition attributes carry this.
• microscope_serial — instrument serial number (Aperture will map serial → user-facing microscope_name like "Snouty" on their side). Serial lives in HardwareSetting metadata.
• imaging_method — normalised imaging modality (confocal, widefield_fluorescence, spinning_disk_confocal, light_sheet). LIF is nearly always confocal (from ATLConfocalSettingDefinition) but derive it, don't hardcode. Use the OME-conventional token set.

Where the values land

Per ADR-0007, values live under per_scene[i].acquisition (or the block name the ADR chooses) in the convert audit AND in each scene entry of inspect().scenes[i]. Missing values → key omitted from the block (no dummy zeros / empty strings). Missing block entirely → key absent from the scene entry.

Implementation notes

• Add extractors to src/zarrmony/metadata/ — new module acquisition.py or extension of lif_channels.py depending on module-size preference at the time.
• Call site: the same _lif_scene_channels code path in api.py walks scene XML for channels — extract acquisition alongside so we only parse once. Thread the returned dict into the scene record.
• Fail-safe like channel + objective extraction: any parse failure returns None and conversion continues without the block (never crash over metadata).
• Bump AUDIT_SCHEMA_VERSION.

Acceptance criteria

• per_scene[i].acquisition populated on a Stellaris LIF fixture with all four fields.
• Same block visible in inspect().scenes[i].
• Missing fields → key omitted; missing block entirely → key absent from scene entry.
• AUDIT_SCHEMA_VERSION bumped; audit.py docstring updated.
• Tests: LIF fixture with all fields, LIF fixture with partial acquisition metadata, LIF fixture with none.
• Post-conversion OME-NGFF validation still passes.
• CHANGELOG entry under a minor version bump.

Blocked by

• ADR-0007: Aperture BigQuery ↔ zarrmony field-mapping contract + instrument-metadata audit shape #60 (ADR-0007 locks the shape)

Follow-up (linked separately)

CZI, ND2, and default (OME-TIFF etc.) reader paths get parallel acquisition + instrument extraction as their own issues, subsuming the earlier #54/#55/#56 objective-only slices.