bigint: small to_str_radix optimization

[ollie27]

Before:

test fac_to_string     ... bench:       1,630 ns/iter (+/- 34)
test fib_to_string     ... bench:         359 ns/iter (+/- 11)
test to_str_radix_02   ... bench:       3,097 ns/iter (+/- 19)
test to_str_radix_08   ... bench:       1,146 ns/iter (+/- 38)
test to_str_radix_10   ... bench:       4,248 ns/iter (+/- 36)
test to_str_radix_16   ... bench:         881 ns/iter (+/- 44)
test to_str_radix_36   ... bench:       8,073 ns/iter (+/- 75)

After:

test fac_to_string     ... bench:       1,492 ns/iter (+/- 20)
test fib_to_string     ... bench:         368 ns/iter (+/- 7)
test to_str_radix_02   ... bench:       2,038 ns/iter (+/- 47)
test to_str_radix_08   ... bench:         812 ns/iter (+/- 9)
test to_str_radix_10   ... bench:       3,919 ns/iter (+/- 40)
test to_str_radix_16   ... bench:         703 ns/iter (+/- 58)
test to_str_radix_36   ... bench:       7,852 ns/iter (+/- 81)

[cuviper]

Huh. I know this table lookup proved faster than the old to_digit, which used a match on ranges. If your single branch is faster still, so be it! Especially for radix <= 10, I'll bet that branch predicts true really well.

CI is failing rust nightly in building rand. :/ rust-random/rand#91 rust-lang/rust#30713

[cuviper]

Nightly vs. rand should be resolved now -- let's give it a whirl...

@homu r+ d2b23d0

[homu]

⌛ Testing commit d2b23d0 with merge e65fd4e...

[homu]

☀️ Test successful - status