Plate: add optional stacking_z_height parameter

[c-reiter]

Summary

Adds an optional stacking_z_height parameter to the Plate class — the vertical pitch (in mm) between two identical plates stacked directly on top of each other (i.e. the height one plate adds to a stack: size_z minus the overlap with the plate below).

This mirrors the existing NestedTipRack.stacking_z_height and gives plates the same notion of stacking pitch that nested tip racks already have. It's required by plate-stacker devices (e.g. the Agilent BenchCel I'm adding in #1109) and defaults to None when unknown, so nothing changes for existing labware.

Per the discussion with @rickwierenga: option (b) — add it as an optional param on Plate rather than a new subclass — split out into its own PR.

Changes

• Plate.__init__: new stacking_z_height: Optional[float] = None parameter, stored on the instance.
• Plate.__eq__: since stacking_z_height is a physical dimension, plates that differ in it are different labware and now compare unequal. __hash__ is restored to Resource's (defining __eq__ otherwise sets it to None), so plates stay hashable, and stacking_z_height is added to __repr__.
• Plate.serialize(): new override so stacking_z_height round-trips through deserialize. This also fixes a pre-existing gap where plate_type was silently dropped on serialization (because Plate had no serialize override, so the constructor arg was never persisted — copy() lost it too).
• ResourceStack: bare plates stacked in the z direction now nest by their stacking pitch. A plate placed directly on another bare plate sinks in by size_z - stacking_z_height, so a stack of N identical plates is size_z + (N - 1) * stacking_z_height tall (both the reported get_size_z() and the child placement). Gated on the value being set, and skipped when the lower plate wears a lid — so existing behavior is unchanged (no library plate sets the value yet). This parallels the lid-nesting ResourceStack already did.
• Docs: documented stacking_z_height (what it is, how to measure it, why it's needed, and the ResourceStack nesting behavior) on the Plate resource page and the ResourceStack notebook, and added it to the example plate definition in the contributor guide.
• Tests: Plate — default-is-None, value-stored, serialize/deserialize round-trip (covering both stacking_z_height and plate_type), equality differs on stacking_z_height, hashability. ResourceStack — two/three plates nest, no-nesting without the value (back-compat), and no nesting onto a lidded plate.

Notes

• No Plate subclasses exist, so adding the keys to serialize() (which deserialize feeds back as constructor kwargs) is safe.
• Backwards compatible: deserializing older JSON without these keys falls back to the constructor defaults; no existing library plate sets stacking_z_height, so ResourceStack behavior is unchanged for all current labware.

[rickwierenga]

thanks!

[BioCam]

Nice!

Should the same be applied to PetriDish?