Add trim/0, ltrim/0 and rtrim/0

[wader]

Trims leading and trailing whitespace.

Was added to jq in jqlang/jq#3056

[wader]

In the jq version of this there is an optimization to not create a new string if there is nothing to trim https://github.com/jqlang/jq/pull/3056/files#diff-bd792def7de6d4b8d70ddc844eabe52f73c264904e263580215219b669a2ca9cR1239-R1242 i tried to follow the rust code but not sure i understood fully how it would behave

[01mf02]

Thanks a lot for your contribution, @wader! It is absolutely exemplary --- updating docs, adding tests, a small implementation. And even linking to your original commit in jq. Great work.

I implemented on top of your work the optimisation that you mentioned, see ec3256a. It relies on the fact that trimming an already trimmed &str returns an identical &str, and we can compare these two in constant time by pointer equality (core::ptr::eq, which considers both the start of the string and the length of the string).

[wader]

Thanks 🙏 and nice you added the optimisation, i haven't researched it but i suspect that most calls to trim will not trim anything. Maybe there are more string functions that could benefit of the same optimisation?

[01mf02]

I just checked that, and ltrimstr as well as rtrimstr already perform this optimisation.
escape_csv and escape_sh are unfortunately no candidates for this optimisation due to the way that str::replace works. The same goes for ascii_downcase and ascii_upcase.

[wader]

👍 now also remembered that i forgot to add tests for some different unicode whitespace codepoints, there is a bunch of different ones, ex a test like this https://github.com/jqlang/jq/blob/master/tests/jq.test#L1363-L1367. I did a quick check and it seems like jq, gojq and jaq all agree on what is whitespace now, ex javascript does not see U+0085 NEXT LINE as whitespace.