Derive the fid receiver count from the receiver count, not from ACQ_dim_desc

[gdevenyi]

Fixes #202.

{"cmd": "1", "conditions": ["#ACQ_dim_desc.__class__.__name__=='str'", "#ACQ_dim_desc=='Spectroscopic'"]},
{"cmd": "1", "conditions": ["#ACQ_dim_desc[0]=='Spectroscopic'"]},
{"cmd": "#PVM_EncNReceivers", "conditions": ["#ACQ_dim_desc[0]!='Spectroscopic'"]}

ACQ_dim_desc describes a dimension as Spatial or Spectroscopic (spec 5.1) and says nothing about receivers. Spec 3.1 defines Nchan as the number of active receiver channels, with no spectroscopy exception — and the premise is false in the corpus: the PV360 PRESS_1H scans are spectroscopic with PVM_EncNReceivers = 4.

channels feeds block_size and acq_length, so a wrong value is a wrong per-scan block length.

The first branch was also dead code. #NAME.accessor expands to self['NAME'].accessor:

$ python -c "from brukerapi.dataset import Dataset; print(Dataset._sub_parameters(None, \"#ACQ_dim_desc.__class__.__name__=='str'\"))"
self['ACQ_dim_desc'].__class__.__name__=='str'

which is the Parameter object's class, always 'GenericParameter'. Only the array-valued branch could fire, i.e. CSI.

Change

Read PVM_EncNReceivers unconditionally — spec 3.1 names it the method-side mirror of the Yes count in ACQ_ReceiverSelect — and fall back to one channel where the method declares none.

No fid in the corpus changes: all 204 keep their channel count, block size and shape.

Scope

This does not by itself make a multi-receiver CSI fid readable — the CSI encoding_space/k_space recipes carry no receiver axis either. With the count right it now fails the storage size check loudly instead of folding the channels into the spectral axis, and no dataset available to the project exercises that layout, so inventing one would be guesswork.

Test

test_a_spectroscopic_acquisition_keeps_its_receiver_count — a synthetic two-receiver CSI experiment (ACQ_dim_desc = Spectroscopic Spatial Spatial, the shape that actually triggered the hard-code). Fails on master.

Suite: 2141 passed, 12 skipped.

[gdevenyi]

Rebased onto #213 (ruff format), so this branch carries that commit until #213 merges — merge #213 first and this diff reduces to its own change. Verified after the rebase: ruff check and ruff format --check clean, full suite green.

Integration check across all eleven conformance PRs merged together: 2164 passed, 12 skipped; corpus load test 3202 → 3207 of 3478 with zero newly broken datasets.