Add BIDS-aware EDF loader with metadata checks

[mnoergaard]

Motivation

• Provide a reusable helper to load EDF files together with BIDS sidecars and ensure channel metadata is harmonized before preprocessing.
• Avoid depending on pandas in the test/runtime path by using the stdlib csv reader for _channels.tsv parsing.

Description

• Add eegprep/workflows/eeg/load.py implementing load_edf_with_bids_metadata which reads the JSON sidecar, reads the channels TSV, loads the EDF via mne, enforces channel count and in-order renaming, maps BIDS type to MNE channel types, checks sampling-rate consistency, assigns a standard montage, and returns (raw, meta, channels).
• Add tests/test_load.py that verifies successful name/type harmonization and that a mismatch in channel counts raises a ValueError using monkeypatch to stub mne.io.read_raw_edf.
• Replace an earlier pandas-based TSV read in tests and loader with csv.DictReader to avoid requiring pandas at test/runtime.

Testing

• Ran pytest -q tests/test_load.py tests/test_preprocess.py which executed the new tests but both were skipped in this environment because mne is not available, resulting in 2 skipped.
• An earlier test run failed due to missing pandas, which prompted switching TSV handling to the stdlib csv and resolving the collection error.

---

Codex Task