feat: allow merge argument for DatasetCollection.add_adatas

[ilan-gold]

To do this we need to disallow Dataset2D in anndata.conat because it does not respect merge arguments...Luckily, these df's are generally small.

It might be time to start thinking about a settings object.

[ilan-gold]

I also slipped this in - I don't think there's much point in outer joining genes if we're just gonna subset them anyway. It seems like we should just subset first and then outer join. I think this is just a distributive law of sets.

[ilan-gold]

See #198 (comment) for a bit more info on a small implication of this, which I think is actually more "correct" anyway

[codecov]

Codecov Report

❌ Patch coverage is 93.10345% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 91.68%. Comparing base (9dc1ba2) to head (28468d5).

Files with missing lines	Patch %	Lines
src/annbatch/io.py	93.10%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #198      +/-   ##
==========================================
- Coverage   93.48%   91.68%   -1.81%     
==========================================
  Files          14       14              
  Lines        1121     1118       -3     
==========================================
- Hits         1048     1025      -23     
- Misses         73       93      +20     

Files with missing lines	Coverage Δ
src/annbatch/io.py	91.77% <93.10%> (+0.44%)	⬆️

... and 3 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ilan-gold]

Ok so here's an interesting problem: by subsetting the gene space first and then concatenating with an outer join, for unique merge, we end up picking up a column that is made to be uniquely identifiable as the other objects (potentially) because its index becomes identical and therefore there is only one option for that column:

# /// script
# requires-python = ">=3.12"
# dependencies = [
#   "anndata",
# ]
# ///


from __future__ import annotations

import numpy as np
import pandas as pd

import anndata as ad

adatas = [
    ad.AnnData(
        X=np.ones((1, 2)),
        obs=pd.DataFrame({"1": ["a"], "2": ["b"]}),
        var=pd.DataFrame(index=["1", "2"], data={"column_1": [1, 2]}),
    ),
    ad.AnnData(
        X=np.ones((1, 2)),
        obs=pd.DataFrame({"1": ["a"], "3": ["b"]}),
        var=pd.DataFrame(index=["1", "3"], data={"column_1": [1, 2]}),
    ),
]
print(ad.concat(adatas, join="outer", merge="unique")[:, ["1"]].var.columns)  # empty
print(
    ad.concat(
        [adata[:, ["1"]] for adata in adatas], join="outer", merge="unique"
    ).var.columns
)  # has column_1

So the question is whether or not we have this. It's a subtle behavioral difference from anndata.concat but I think it is actually correct for our use-case.