Fix subepoch fallback channel inference for Intan and Open Ephys

[yoshihito-saito]

This fixes subepoch binary fallback loading when no concatenated .dat file is present.

Root cause:

• Open Ephys continuous.dat files can include extra trailing channels beyond session.extracellular.nChannels.
• The fallback channel inference relied on MergePoints.timestamps_samples sample counts.
• MergePoints sample ranges behave like inclusive ranges, causing an off-by-one sample mismatch and non-integer inferred channel counts.

Changes:

• Infer file channel count using expectedSamples, expectedSamples + 1, and expectedSamples - 1, choosing the one consistent with file size.
• Preserve reading only 1:session.extracellular.nChannels when fallback files contain extra channels.
• Resolve MergePoints.foldernames that contain full Linux paths by matching the local epoch folder name.

Validated on:

• Open Ephys: RM018_day39_260620
  • continuous.dat: 200 channels
  • session: 192 channels
  • extra ADC channels excluded
• Intan: sake_day10
  • amplifier.dat: 128 channels
  • session: 128 channels

Tests:

• tests/test_getWaveformsFromDat_mergepoints.m
• Minimal getWaveformsFromDat fallback check on both real sessions