MCMICRO reader does not support new nf-core/mcmicro output formats

We recently completed a port of MCMICRO to nf-core (https://nf-co.re/mcmicro/2.0.0). There are minor changes to expected output formats, and the current spatialdata_io.readers.mcmicro reader — written against the original labsyspharm/mcmicro layout — does not read nf-core output.

What changed between the two pipelines:

Gotchas found while testing against real nf-core/mcmicro output that will need to be considered:

• Segmentation mask filenames are tool-specific, not a uniform _mask suffix — e.g. 1_cp_masks.tif (cellpose) vs exemplar-002_1_mask.tif (mesmer).
• The segmenter directory name differs between segmentation/ and quantification/mcquant/ (e.g. deepcell_mesmer vs mesmer), so tables can't be linked to labels by an exact directory-name match.
• With background subtraction, the registration image and the quantification tables carry different marker sets (e.g. 40 image channels vs 37 table markers).
• TMA core numbers must be parsed carefully — the sample name itself can contain digits (exemplar-002_1_mask → core 1, not 002) — and tma_dearray/ also contains TMA_MAP.tif, which is not a core.
• The prelude/markers_markersheet_mqc.tsv file is a long-format MultiQC validation report, not a usable wide marker sheet.

Proposal

Update the mcmicro reader to auto-detect the pipeline (with an explicit pipeline= override) and handle both layouts, including multiple samples and segmenters, in WSI and TMA modes. Backwards compatibility with labsyspharm output is preserved.

Start in spatialdata_io.readers.mcmicro and compare its assumptions with the two output layouts listed in the issue. Trace how registration, segmentation, quantification, markers, and TMA cores are discovered, then define validation cases for multiple samples and segmenters, pipeline auto-detection, and the explicit override. Done means both labsyspharm and nf-core outputs remain readable, including WSI and TMA cases.