Refactor samples.tsv schema: explicit input_type column and output_bids_dataset support

[Copilot]

Workflow selection was implicitly derived from acq substrings (e.g., {acq,[a-zA-Z0-9]*blaze[a-zA-Z0-9]*}), making it fragile and unintuitive. A single root output path also prevented multi-dataset workflows.

Schema changes

samples.tsv gains two new columns:

Column	Values	Purpose
input_type	blaze | prestitched | imaris	Explicit workflow selection per sample
output_bids_dataset	path (optional)	Override output root per-run

subject  sample  acq   input_type  output_bids_dataset  stain_0  sample_path
mouse1   brain   4x    blaze       /data/bids_out        Lectin   /data/raw
mouse2   brain   4x    prestitched                       DAPI     /data/raw2

Routing refactor

• Replaced all inline {acq,[a-zA-Z0-9]*blaze[a-zA-Z0-9]*} wildcard patterns with wildcard_constraints blocks driven by the samples table
• Added get_acq_constraint(input_type_str) — generates a per-type regex from actual acq values in samples; returns a never-matching regex ((?!x)x) when no samples use that type
• get_output_ome_zarr() no longer takes an acq_type argument; each calling rule declares its own constraint

rule zarr_to_ome_zarr:
    wildcard_constraints:
        acq=get_acq_constraint("blaze"),
    output:
        **get_output_ome_zarr(),

output_bids_dataset support

• If the column is present and all rows share one non-empty value, it overrides the global root config before rules load
• Multiple distinct values raise a clear ValueError with a message pointing to future per-sample routing support

CLI

• Adds --input-type {blaze,prestitched,imaris} (default: blaze) so the single-sample run.py path also sets the column explicitly

Bug fix

imaris_to_ome_zarr was referencing prestitched_to_metadata output; corrected to imaris_to_metadata.