perf: improve performance of string repeat

[andygrove]

Which issue does this PR close?

N/A

Rationale for this change

Use a re-usable string buffer instead of allocating a new string for each input value.

Benchmark	Main (µs)	Optimized (µs)	Improvement
size=1024, repeat=3
repeat_string_view	76.51	70.14	-8.3%
repeat_string	78.63	71.41	-9.2%
repeat_large_string	76.40	71.08	-7.0%
size=1024, repeat=30
repeat_string_view	109.02	93.51	-14.2%
repeat_string	108.46	92.12	-15.1%
repeat_large_string	105.99	91.66	-13.5%
size=4096, repeat=3
repeat_string_view	139.44	113.95	-18.3%
repeat_string	133.62	112.25	-16.0%
repeat_large_string	131.94	108.41	-17.8%
size=4096, repeat=30
repeat_string_view	251.77	193.95	-23.0%
repeat_string	250.58	191.86	-23.4%
repeat_large_string	248.88	188.43	-24.3%
overflow tests
size=1024	58.14	58.02	~0% (no change)
size=4096	58.26	58.08	~0% (no change)

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

[Jefffrey]

We could even try creating the values/offsets/null buffers manually, in order to copy the strings directly into the values buffer instead of the intermediate buffer to skip builder API completely, but perhaps gets too into the weeds