feat(snapshots): add snapshot fixtures, remove pandas fixture

[wpbonelli]

• add numpy array snapshot extensions and fixtures with syrupy, moved from flopy
• remove use_pandas fixture, originally meant for use in flopy tests for pandas support but never used and probably not needed, as flopy use_pandas flags will be removed eventually

[wpbonelli]

@mwtoews I was mainly wondering what you thought of the optional import decision here. But I think it's probably best to keep using import_optional_dependency(), will update to do so unless you disagree

[mwtoews]

@wpbonelli no strong opinions so far, I'm just finding time to look further at the methods. One further idea is to also support compressed array ".npz" using numpy.savez_compressed.

[wpbonelli]

Thanks @mwtoews for the suggestion, savez_compressed() also allows snapshots of multiple arrays at once. I pushed a commit drafting support for multiple arrays e.g.

def my_test(multi_array_snapshot):
    ...
    arrays = {
        "head": ...
        "budget": ...
    }
    assert arrays == multi_array_snapshot

which is consistent with np.load() returning a dict for multi-array .npz files.

[wpbonelli]

Looks like with npz files we get inconsistent binary encodings. Strange that it's only npz files and not npy files.

[wpbonelli]

I will merge as-is so we can use this for PRT tests, but would be ideal to support snapshots of multiple arrays eventually.