Align documentation and workflow examples with current behavior

[araikes]

The current documentation correctly describes the basic single-scalar HDF5 workflows, but it does not fully or consistently document other functions or complete examples. Specifically, the documentation should be extended to explain:

• source_mask_file is currently required for NIfTI cohorts, even when the group mask is reused for every subject.
• Wide-format CSVs are supported with --scalar-columns.
• Wide and long formats have different default output-splitting behavior.
• Multiple scalars can be combined into one output or split into separate outputs.
• CIFTI support includes dscalar, pscalar, and pconn workflows.
• TileDB is a supported conversion backend with a distinct output layout.
• Actual compression defaults and backend-specific compression options.
• Actual HDF5 and TileDB metadata locations.
• S3 installation, usage, worker configuration, and modality limitations.
• export-results reads ModelArray statistical result groups rather than directly exporting imported scalar datasets.
• Existing-file overwrite behavior for each exporter.
• The expected result matrix shape and result-name lookup behavior.