Use String#byteindex instead of String#index

[shugo]

Because String#index takes O(N), where N is the second argument, while String#byteindex takes O(1).

[shugo]

We should wait until Ruby 3.1 gets EOL.

[nevans]

@shugo I added it to a new v0.6 milestone. 🙂

[admsev]

@gooroodev WDYT?

[nevans]

@admsev@gooroodev please don't. It's obnoxious and not very helpful.

[shugo]

I remember now that the argument of ResponseParser#parse is usually ASCII-8BIT, so this change doesn't matter unless ResponseParser#parse is directly used with other encodings.

[nevans]

@shugo, as you said, our inputs are always ASCII-8BIT, so this isn't really needed. I do want it for semantic correctness. but I suspect #byteoffset vs #end will actually hurt performance due to the Array allocation. So this PR will need to wait until the min required ruby is 3.4 (after 2027-04-01) before we can use #byteend. (Alternatively, we could rewrite it with if MatchData.instance_methods.include?(:byteend), but I think that's maybe more hassle than is justified.)