Use String.repeat() to optimize several String methods

[MewPurPur]

kleonc and I optimized String.repeat() in #64489, so I wondered if there there is any engine code that is using iterative addition instead of this method. Nice surprise though, I found that even some user-exposed String methods could be optimized this way, namely pad_zeros(), pad_decimals(), rpad(), and lpad(). The latter two are even used in String::sprintf() - I have no idea what sprintf does, but it looks important!

The optimization should always result in performance gain, though I haven't benchmarked it extensively.

Edit: It wasn't always a performance gain. It's just that the old repeat() was really slow, but the new one is still slower than a manual loop if it has 1 or 2 iterations. For a single iteration, the performance reduction was 15%, which I wasn't happy with, so I altered String::repeat() a bit to return the string directly if p_count is 1. This doesn't affect performance for other cases in any noticeable way.

Testing new vs. old rpad() after that:

Test	Performance boost/reduction
0 repeats	+20%
1 repeat	+8%
2 repeats	-3%
3 repeats	+12%
4 repeats	+30%
10 repeats	+150%
16 repeats	+420%

RegEx used: for \(.*\) {\n((\t)*).* \+= .*;\n((\t)*)} and lots of manual sifting. A few I found with slightly different regexes too.

[MewPurPur]

Instances I found but I'm not sure if I should edit: doc_tools.cpp, bvh_debug.inc, gltf_document.cpp

I'll push these too, but I don't understand what some of the files are for, so I'll appreciate a review to make sure I'm not going over the top

[AThousandShips]

Makes me wonder if we should add a String(count, char) constructor for convenience

[lawnjelly]

I'll push these too, but I don't understand what some of the files are for, so I'll appreciate a review to make sure I'm not going over the top

Generally it looks pretty easy to add and self-explanatory, so I wouldn't worry too much about over adding it (whether it will make any difference is another matter 😁 ).

Seeing what appears when profiling is usually the gold standard for optimization, but on the other hand text output tends to occur in just a few frames, rather than continually, which can make it harder to identify via profiling.

[MewPurPur]

Not 100% safe and not a bugfix. Makes sense to bump the milestone.

[kleonc]

LGTM besides the one potential change of the behavior.

[adamscott]

Looks good to me, once @kleonc comments are addressed.

[mhilbrunner]

LGTM. Also makes the code nicer to read IMO :)

[clayjohn]

Great work! Thanks!