S3FileSystem.info(version_id=...) returns cached metadata of another version #932

Problem

S3FileSystem.info(path, version_id=...) returns cached metadata of a different version. After info(p, version_id="v1"), a call to info(p, version_id="v2") returns v1's size, ETag and version_id instead of v2's.

Expected: an explicit version_id is honored, either by matching it against the cached entry or by bypassing the cache.

From the code on master (775874c):

• _head_object() caches the result under the path without the version (self.dircache[path] = file) and returns a cached entry whenever path in self.dircache, without comparing the requested version_id.
• In info(), a cached entry is returned as is unless the filesystem is version aware, the call has no version_id, and the cached entry has none either. Calls with a different explicit version_id therefore get the cached entry.

Found by code reading during the review of #928 (not reproduced live yet).

Reproduction

from pyathena import connect
from pyathena.filesystem.s3 import S3FileSystem

fs = S3FileSystem(connect(), version_aware=True, skip_instance_cache=True)
path = "s3://VERSIONED_BUCKET/tmp/key"
fs.pipe(path, b"1")
v1 = fs.info(path, refresh=True)["version_id"]
fs.pipe(path, b"22")
v2 = fs.info(path, refresh=True)["version_id"]

fs.invalidate_cache()
print(fs.info(path, version_id=v1)["size"])  # 1
print(fs.info(path, version_id=v2)["size"])  # expected 2, returns 1 from the cache

Environment

• PyAthena master 775874c, S3FileSystem with version_aware=True.

Proposed fix (optional)

When a version_id is requested (as an argument or in the path), use the cached entry only if its version_id matches; otherwise head the object with that version. Add a test that reads two versions of the same object through one instance.