ARROW-11243: [C++] Recognize time types in CSV files

[pitrou]

• Allow reading CSV columns as time32 and time64
• Automatically infer "hh:mm" and "hh:mm:ss" as time32[s]

[github-actions]

https://issues.apache.org/jira/browse/ARROW-11243

[pitrou]

@ursabot please benchmark

[ursabot]

Benchmark runs are scheduled for baseline = 169b249 and contender = 25b5c1e445f04b53a0beb40afabb3e8012c31f50. Results will be available as each benchmark for each run completes.
Conbench compare runs links:
[Finished ⬇️0.0% ⬆️0.0%] ec2-t3-xlarge-us-east-2 (mimalloc)
[Finished ⬇️0.0% ⬆️1.53%] ursa-i9-9960x (mimalloc)
[Finished ⬇️0.91% ⬆️0.1%] ursa-thinkcentre-m75q (mimalloc)
Supported benchmarks:
ursa-i9-9960x: langs = Python, R, JavaScript
ursa-thinkcentre-m75q: langs = C++, Java
ec2-t3-xlarge-us-east-2: cloud = True

[bkietz]

This looks good, I'd just like the unit tests for inference to be expanded a bit

[nealrichardson]

Seems reasonable to me, just one question about time64

[nealrichardson]

Should there also be an inferring test that results in time64?

[pitrou]

Currently, there is no inference to time64. Should there be one (for times with nanosecond precision perhaps)?

[nealrichardson]

Ah I see, I was looking at the wrong case statement, you're right. Maybe there should be one to catch any times with fractional seconds, though I don't feel strongly about it since you can explicitly declare the type/schema you want.

[pitrou]

(rebased)

[westonpace]

This could be a future JIRA or "wait until it's asked for" but technically ISO-8601 allows the omission of the colons if you add a leading T but I've never seen it in practice...

Thh
Thhmm
Thhmmss
Thhmmss.sss

[pitrou]

@westonpace I'd rather wait for someone to request it.