[WIP] Use CompactString as backing storage for Steel strings.

[wmedrano]

• Use CompactString.
• Benchmark.

[wmedrano]

Here are the benchmark results I got.

TLDR: Seems a little bit better, but maybe not enough. May run again since results may be noisy. May also look into some string benchmarks.

Results

• Ran the same benchmarks as CI.
• Used diff --side-by-side --suppress-common-lines before.txt after.txt to filter output.

Before	After	Diff
test range-big ... bench: 82825 ns/iter (+/- 292)	test range-big ... bench: 82320 ns/iter (+/- 334)	-1%
test map-big ... bench: 465220 ns/iter (+/- 1583)	test map-big ... bench: 433713 ns/iter (+/- 1897)	-7%
test transducer-map ... bench: 1144534 ns/iter (+/- 3158)	test transducer-map ... bench: 1071335 ns/iter (+/- 5610)	-6%
test filter-big ... bench: 340884 ns/iter (+/- 3501)	test filter-big ... bench: 331847 ns/iter (+/- 4115)	-3%
test ten-thousand-iterations ... bench: 668004 ns/iter (	test ten-thousand-iterations ... bench: 600138 ns/iter (	-10%
test ten-thousand-iterations-letrec ... bench: 1173667 ns	test ten-thousand-iterations-letrec ... bench: 1120359 ns	-5%
test trie-sort-without-optimizations ... bench: 326192 n	test trie-sort-without-optimizations ... bench: 335860 n	+3%
test fib-28/fib-28 ... bench: 47478721 ns/iter (+/- 266770	test fib-28/fib-28 ... bench: 44984746 ns/iter (+/- 168274	-5%
test thread-creation/thread-creation ... bench: 1022818 n	test thread-creation/thread-creation ... bench: 1034177 n	+1%
test engine-creation ... bench: 18445246 ns/iter (+/- 1228	test engine-creation ... bench: 18327267 ns/iter (+/- 1816	-1%
test register-fn ... bench: 138 ns/iter (+/- 0)	test register-fn ... bench: 146 ns/iter (+/- 0)	+6%
test multiple-transducers ... bench: 6524 ns/iter (+/-	test multiple-transducers ... bench: 6359 ns/iter (+/-	-3%
bench: 336427485 ns/iter (+/- 4025646)	bench: 331793506 ns/iter (+/- 3159865)	-1%
test ackermann-3-3 ... bench: 272159 ns/iter (+/- 762)	test ackermann-3-3 ... bench: 232964 ns/iter (+/- 1183)	-14%