datetime: redo for full features

[gigagrug]

Description

1. Add Unixepoch modifier
2. Add JulianDay modifier
3. Add Auto modifier
4. Fix Localtime and UTC shouldn't double convert
5. fix err bug: returns "" needs to return null SELECT typeof(strftime('%Y', 'not-a-valid-date'));
6. add support for decimal month and year calculations (on top of jussi's added support for days, hours, min, sec)
7. fix the bugs @LeMikaelF found in my code from Full datetime support #4104.
   Support modifiers for UnixEpoch and JulianDay functions #602

Motivation and context

I originally had this pr: #4104
But then I saw: #4345 that did things better.
However although my code uses some code in the PR I couldn't add all the code without breaking compatibility. Specifically the faster parser.

There is also this PR: #4331
Storing date in struct is closer to how SQLite handles date but I personally found too much reliance on the Chronos create a lot of bugs with compatibility.

Description of AI Usage

I used Google Gemini Pro.

1. have it generate tests first.
2. then verify tests in sqlite and fix the mistakes.
3. generate code based on tests.
4. run tests.
5. tell Gemini the failed tests.

[gigagrug]

@LeMikaelF I know you're busy but can you help me with this. I have no idea how my code would cause this error:
FAIL [ 0.628s] ( 999/1605) turso_core mvcc::database::tests::test_cursor_with_btree_and_mvcc_fuzz

[LeMikaelF]

@LeMikaelF I know you're busy but can you help me with this. I have no idea how my code would cause this error: FAIL [ 0.628s] ( 999/1605) turso_core mvcc::database::tests::test_cursor_with_btree_and_mvcc_fuzz

You were just (unlucky) enough that your CI run found a failing seed in one of the fuzz tests. I've opened an issue #4371 and restarted the CI action. It should be good now.

Don't hesitate to tag us in for things like this.