fix: starts_with_ansi always returned 1 and had inverted logic

[rwe]

$ansi_color_regex is defined as (…)? and so /^$ansi_color_regex/ will always match. This PR factors the ANSI-sequence part of the regex itself and checks (non-optionally) against that.

Additionally, the logic checking starts_with_ansi was inverted, likely to compensate for the above error, but as a side effect forcing diff-so-fancy always to color lines unnecessarily.

[scottchiefbaker]

Interesting... good catch!