Simplify and speed up timestamp parsing

[akx]

As mentioned in #2972 (comment):

_parse_timestamp_with_tzinfo() already attempts to do fromtimestamp parsing; not much point in doing that work twice for timestamp-esque strings (and failing in the cases described in #2972).

In the subsequent commits, this PR also cleans up remaining seemingly Useless Use of tzlocal.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

All modified and coverable lines are covered by tests ✅

Please upload report for BASE (develop@976b894). Learn more about missing BASE report.
Report is 177 commits behind head on develop.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #3063   +/-   ##
==========================================
  Coverage           ?   93.06%           
==========================================
  Files              ?       66           
  Lines              ?    14472           
  Branches           ?        0           
==========================================
  Hits               ?    13468           
  Misses             ?     1004           
  Partials           ?        0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[akx]

@nateprewitt I have some tests you seem to have added for #1987 failing with this (hence a draft), but you might have some context to help here: Why do we need tzlocal or tzwinlocal in the first place?

parse_timestamp doesn't necessarily guarantee the returned datetime should have any tzinfo attached, though the tests seem to always assume it's UTC, and practically, parsing any timestamp off the wire does prescribe a timezone in some way (GMT for RFC822, Z for ISO8601).

Might I be right in guessing the tzlocal/tzwinlocal stuff could actually be all removed in favor of _epoch_seconds_to_datetime's arithmetics?

[akx]

(Gentle review nudge, @nateprewitt? 😄)

[akx]

Another nudge, @nateprewitt... I noticed you'd said

we still have some concerns around the original windows issue that lead to that PR. It's likely safe to clean up as you've suggested but we'll need to do some manual testing first.

in #3062 – anything I can do to help things along?

[akx]

Sorry for Yet Another Ping, @nateprewitt 😅

[akx]

Another ping... @nateprewitt?

[akx]

Rebased post conflicts from #3206...

[akx]

I decided to just add the optimization commit I'd mentioned in a previous comment in here, because why not – after all this PR has been brewing for over a year so might as well.

Benchmark results for parsing timestamps found in the test suite comparing 05538ce and this PR, showing a mean OPS improvement of 16.75×:

Name (time in ms)	OPS	Min	Max	Mean	StdDev	Median	IQR	Outliers	Rounds
test_benchmark_parse_timestamp[develop]	44.7751	20.0351	25.8432	22.3338	1.7029	22.2602	2.9707	86;0	211
test_benchmark_parse_timestamp[this PR]	750.3208	1.1025	1.6340	1.3328	0.1388	1.2426	0.2904	929;0	2541