[ENH] Improve BIDS compatibility of PetLinear CAPS output

[NicolasGensollen]

The general idea of this PR is to improve the BIDS compliance of the PETLinear pipeline.

The current output looks like this:

caps
└── subjects
    ├── sub-01
    │   ├── ses-M000
    │   │   ├── pet_linear
    │   │   │   ├── sub-01_ses-M000_trc-18FFDG_pet_space-MNI152NLin2009cSym_desc-Crop_res-1x1x1_suvr-pons_pet.nii.gz
    │   │   │   └── sub-01_ses-M000_trc-18FFDG_pet_space-T1w_rigid.mat

Where the suffix pet appears after the tracer while it shouldn't.

This PR proposes to:

• Refactor the PETLinear utility function rename_into_caps
• Improve documentation and variable names
• Fix the issue described above (which came from the fact that the suffix wasn't stripped from the BIDS source file name used to get the source entities)
• Update the PETLinear pipeline accordingly
• Add unit tests

Since we are changing the output, this PR needs the following data PR : https://github.com/aramis-lab/clinica_data_ci/pull/47

[ghisvail]

LGTM, with a few comments for posterity.

I have got also comments on the data PR.

[ghisvail]

Just a comment that we could achieve further compliance by adopting BEP014 which specifies a naming scheme for transformation files, i.e.

<participant_id>/
    <session_id>/
        xfm/
            <participant_id>_<session_id>_from-T1w_to-MNI152NLin2009cSym_mode-image_xfm.mat

[ghisvail]

It would probably require more surgery to t1-linear and pet-linear though, so it's probably better to leave it for another time. It was just a reminder.

[NicolasGensollen]

I agree, let's keep that for another PR.
Thanks for the reminder !