Avoid parse errors of datetime.isoformat strings

[toshihikoyanase]

Motivation

@nzw0301 found that CI failed due to the error of JournalStorage. They said the dumped strings of datetime_start (or datetime_complete) could lack microseconds but datetime_from_isoformat function always expects them.

import datetime
datetime.datetime(2022, 9, 29, 2, 27, 32, 0).isoformat()
# '2022-09-29T02:27:32'

optuna/optuna/storages/_journal/storage.py

Lines 48 to 50 in 31da4d0

	def datetime_from_isoformat(datetime_str: str) -> datetime.datetime:
	# TODO(wattlebirdaz): Use datetime.fromisoformat after dropped Python 3.6 support.
	return datetime.datetime.strptime(datetime_str, "%Y-%m-%dT%H:%M:%S.%f")

@hvy suggested workarounds in https://bugs.python.org/issue40076, and this PR employs the timespec argument of datetime.isoformat to always show microseconds.

Description of the changes

• Specify timespec="microseconds" to datetime.isoformat()
• Add test case to handle template trials with datetime_start/datetime_complete whose microseconds values are zero.

[toshihikoyanase]

@wattlebirdaz @c-bata Could you review this PR as authors of JournalStorage, please?

[codecov-commenter]

Codecov Report

Merging #4025 (cee4443) into master (31da4d0) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #4025   +/-   ##
=======================================
  Coverage   90.19%   90.19%           
=======================================
  Files         160      160           
  Lines       12580    12580           
=======================================
  Hits        11347    11347           
  Misses       1233     1233           

Impacted Files	Coverage Δ
optuna/storages/_journal/storage.py	96.50% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[rotaki]

Thank you for finding out the bug and fixing it. LGTM.

[c-bata]

Thank you for your fix! LGTM.