[WIP] add macsima reader

[berombau]

Redo of previous work in #156.

This tries to solve #155.

Lung_adc_demo

from spatialdata_io import macsima
sdata = macsima('./Lung_adc_demo', subset=1000)
sdata

Note that now the cycle information is added to a table and that the coordinate systems are named after the dataset.

SpatialData object
├── Images
│     └── 'Lung_adc_demo_image': DataTree[cyx] (116, 1000, 1000), (116, 500, 500)
└── Tables
      └── 'Lung_adc_demo_table': AnnData (0, 116)
with coordinate systems:
    ▸ 'Lung_adc_demo', with elements:
        Lung_adc_demo_image (Images)

For the LiverCellAtlas data, it stores all middle cycle nuclei channels in a different image and table.

SpatialData object
├── Images
│     ├── 'HumanLiverH36_image': DataTree[cyx] (102, 5772, 6621), (102, 2886, 3310), (102, 1443, 1655), (102, 721, 827)
│     └── 'HumanLiverH36_nuclei_image': DataTree[cyx] (50, 5772, 6621), (50, 2886, 3310), (50, 1443, 1655), (50, 721, 827)
└── Tables
      ├── 'HumanLiverH36_nuclei_table': AnnData (0, 50)
      └── 'HumanLiverH36_table': AnnData (0, 102)
with coordinate systems:
    ▸ 'HumanLiverH36', with elements:
        HumanLiverH36_image (Images), HumanLiverH36_nuclei_image (Images)

Parsing of raw data should be another PR. It differs in the handling of metadata for channels and cycles from https://gustaveroussy.github.io/sopa/api/io/#sopa.io.macsima @quentinblampey.

The channel naming is different (R{cycle} {marker} instead of {marker}{ (duplicate_number if duplicate markers)). I like both styles, so I'll try to add later style as well and add a parameter.