fix(s3): correct S3 block size error messages to match checks

[Jah-yee]

Summary

Three S3 block size error messages did not match the actual validation logic:

1. s3.py:1278 & s3_async.py:284: The check uses block_size < MIN_PART_SIZE or > MAX_PART_SIZE (inclusive bounds), but the message said "greater than 5MiB and less than 5GiB" (exclusive). Updated to "at least 5 MiB and at most 5 GiB" to match the actual inclusive bounds.

2. s3.py:2325: The message showed the byte count (5242880) with an "MB" suffix, which is confusing. Updated to "5 MiB (5242880 bytes)" to make the units clear.

Changes

• pyathena/filesystem/s3.py: Fix two ValueError messages
• pyathena/filesystem/s3_async.py: Fix one ValueError message

Reproduction

from unittest.mock import MagicMock
from pyathena.filesystem.s3 import S3FileSystem

fs = S3FileSystem(connection=MagicMock(), skip_instance_cache=True)
fs.open("s3://bucket/key", "wb", block_size=1024)
# Before: ValueError: Block size must be >= 5242880MB.
# After:  ValueError: Block size must be >= 5 MiB (5242880 bytes).

Fixes #926.

[laughingman7743]

Thank you for the fix, @Jah-yee!

#926 was already being fixed together with the related S3File write validation (#976 and #952), which changes the same checks and messages, in #992. So I'm closing this PR in favor of that one. Thanks again for taking the time to look into it.