Figure out sample merging

[nleroy917]

Sample merging is quite simple. And we can also probably easily implement it with pandas group_by, the only issue is aggregating multiple samples into lists. It doesn't seem to want to work the way I expected; it makes everything a list.

[khoroshevskyi]

Problems in duplicated sample_names in sample_table:

1. Caveat: Polars columns must have a single type — a column can't be i32 for some rows and List<i32> for others. So in practice, the cleanest design is to always use List<T> for columns that can ever have duplicates, and let downstream code handle single-element lists. The "smart unwrap" works if you process per-column across the whole dataset, not per-row.
2. If we want to merge samples, Polars wants to convert everything in lists, even columns that have duplicated values. To avoid it, we would have to write custom function that would review each column, and decide if value is or is not duplicated.

Based on this information, we won't implement this functionality. This functionality will be available only in functions: get_sample and get_subsample

Subsample tables are implemented, and work as designed in specifications