Add a feature flag to only count LF and CRLF as line breaks.

[cessen]

Most other software does not conform to the Unicode spec when it comes to line breaks. This (understandably) gives people pause when deciding whether to use Ropey or not, because although Ropey is "correct" in the way it counts line breaks, it's not compatible with other software if any line breaks other than LF/CRLF are present in a file.

We can solve this with a feature flag that implements the simpler, non-Unicode-conformant line break counting that is common to most other software.

[archseer]

Also related to #43

[cessen]

Oh! Right. I feel silly now. Basically a duplicate.

[cessen]

I split Ropey's str_utils module off into a separate crate, str_indices, which now also implements two additional ways of line counting:

• Only recognizing LF.
• Only recognizing LF and CR.

That was the "hard" part. The rest should be relatively straightforward:

1. Add the feature flags for these two alternative approaches, which just swap out which functions are imported from str_indices.
2. Update all of the line-based unit tests and property tests to verify against the output of those imported functions rather than hard-coded expected results.

Although straightforward, there are a lot of unit tests to update. So it's a bit of a project still.

[cessen]

Added in 38531e083f8cf5c5b1aeb328361895fb5a9b867c.

There are two new feature flags:

• cr_lines, which enables carriage return line break handling.
• unicode_lines, which enables full Unicode line break handling.

Both features are enabled by default for backwards compatibility, but can be disabled to strip things down. Note that unicode_lines implies cr_lines, since the former is a superset of the latter.

[cessen]

On second thought: re-opening, just to remind myself to add proper documentation for this before the next release.

[cessen]

Documentation done.