String#chars is slow

[lopopolo]

For both ASCII-only UTF-8 Strings and Binary Strings.

artichoke@0dd6a602bac4:/$ time artichoke -e 'x = "abc" * 10_000_000' -e 'puts x.chars.length'
30000000

real    0m5.372s
user    0m4.749s
sys     0m0.622s
artichoke@0dd6a602bac4:/$ time artichoke -e 'x = "abc".b * 10_000_000' -e 'puts x.chars.length'
30000000

real    0m5.405s
user    0m4.765s
sys     0m0.641s

On MRI these operations are twice as fast:

$ time ruby -e 'x = "abc".b * 10_000_000' -e 'puts x.chars.length'
30000000
ruby -e 'x = "abc".b * 10_000_000' -e 'puts x.chars.length'  2.75s user 1.83s system 99% cpu 4.585 total

For cases where a String is ASCII compatible, this operation doesn't need to go through the iterator (and especially doesn't need to go through UTF-8 decoding). It should be quicker to do a copy and map over individual bytes to allocate the many String objects.

This probably requires a new API on spinoso_string::String and for the split between owned and borrowed encoded strings to land for binary and ascii encodings.