Support converting large dates (i.e. +10999-12-31) from string to Date32

[phillipleblanc]

Which issue does this PR close?

• Closes Support casting strings to Date32 that contain large dates #7073

Rationale for this change

Support for casting large dates from string to Date32.

What changes are included in this PR?

Extend the parse_date method, which is used in the impl Parser for Date32Type, to handle dates which are prefixed with + or -. If the date is not prefixed with + or -, the existing logic is used unmodified.

This code isn't as optimized as the code for processing more common date formats - but given that these extended dates are relatively rare in practice, I don't think it matters all that much.

Are there any user-facing changes?

Aside from the desired fix, no.

[alamb]

Thank you @phillipleblanc -- this looks readonable to me. I only think the PR needs a few more tests and we can merge.

I am sure we could make parsing dates like this faster but we can do that type of optimization as a follow up.

I am also running the cast benchmarks just to be sure this doesn't accidentally introduce a regression and will post the results to this PR

[alamb]

Without the changes in this PR this code fails like this (so the test does cover it)

called Result::unwrap() on an Err value: CastError("Cannot cast string '+10999-12-31' to value of Date32 type")

[alamb]

Could you also please add tests for:

            Some("0010-02-28"),

(to show that the negative value actually results in a different parsed value)

"0000-00-00"
"-0000-01-01"
"-0001-01-01"

(for boundary cases)

Also a test that shows trying to parse Some("10999-12-31"), (more than 4 year digits) correctly results in an error?

[alamb]

It was very non clear to me at first that the sign results in different rules. Can we add some comments to clarify? Something like:

Also, can you include a liml to the reference were you got that sentenece?

Suggested change

	// If the date has an extended (signed) year such as "+10999-12-31" or "-0012-05-06"
	// If the date has an extended (signed) year such as "+10999-12-31" or "-0012-05-06"
	//
	// According to ISO 8601, years have:
	// Four digits or more for the year. Years in the range 0000 to 9999 will be pre-padded by
	// zero to ensure four digits. Years outside that range will have a prefixed positive or negative symbol.

[alamb]

++ critcmp main phillip_250205-handle-large-dates
group         main                                   phillip_250205-handle-large-dates
-----         ----                                   ---------------------------------
2020-09-08    1.00     21.5±0.05ns        ? ?/sec    1.04     22.2±0.05ns        ? ?/sec
2020-09-8     1.01     19.0±0.08ns        ? ?/sec    1.00     18.8±0.03ns        ? ?/sec
2020-9-08     1.00     18.6±0.04ns        ? ?/sec    1.04     19.4±0.15ns        ? ?/sec
2020-9-8      1.00     17.4±0.02ns        ? ?/sec    1.01     17.5±0.02ns        ? ?/sec

Seems ok to me