cellbender allows setting learning rate per sample.

[hl-xue]

Allow setting a learning rate per sample in pipeline_cellbender.py.

• If a learning_rate field is defined at the sample level, its value will be used.
• Otherwise, the global learning_rate specified at the cellbender level will apply.

Example YAML with sample-specific learning rates:

cellbender:
    learning_rate: 1e-4
    # ... other parameters

samples:
    sample1:
        expected_cells: 1600
        total_droplets_included: 25000
        learning_rate: 1e-5 # Overrides global
    sample2:
        expected_cells: 1500
        total_droplets_included: 20000
        # no learning rate => uses global 1e-4
    # ... other samples