S3FileSystem.find(maxdepth=..., prefix=...) ignores the depth limit when prefix contains a slash

[laughingman7743]

Problem

S3FileSystem.find(path, maxdepth=n, prefix=...) does not limit the depth when prefix contains /.
With maxdepth, _find() lists Prefix=f"{key}/{prefix}" with Delimiter="/" (pyathena/filesystem/s3.py:733 on master e0e85da) and treats the returned entries as the first level below path.
S3 groups by the delimiter only after the full prefix, so a prefix such as sub/deep/ makes the "first level" entries of path actually lie two levels down, and every further level is shifted the same way.

fsspec's glob() passes only a slash-free filename stem as prefix, so this affects direct find(..., prefix=...) callers.

Found by the independent review of #956 (#933).

Reproduction

fs.touch("s3://BUCKET/d/sub/deep/x")
fs.find("s3://BUCKET/d", maxdepth=1, prefix="sub/deep/")
# expected: [] (nothing directly under d/ starts with "sub/deep/")
# actual: ['BUCKET/d/sub/deep/x']

Environment

• PyAthena master e0e85da (maxdepth and prefix together since v3.15.0).

Proposed fix (optional)

Decide what prefix with a / should mean together with maxdepth (reject it, or count the levels from path), and document it in the find() docstring.