Draft fix for duplicate leading double-asterisk, and **/ edge cases

[JonjonHays]

Hi there @cpburnz! First off, thanks so much for this library -- it's been incredibly useful to my team! I found a couple edge-case-y bugs. There's two issues I found, and the test cases I added are mostly self documenting as far as describing the bugs. I've added a (possibly hacky) fix for the first one, "duplicate leading double-asterisk". Basically the change I made just trims duplicate **/s ...The second one is for the **/ pattern. This one will likely require changes in the _translate_segment_glob(seg) method (I'm not really sure, actually). Anyways, I think we will probably deal with these bugs within our own code, but I figured I'd at least provide a couple test cases and a launch point instead of just filing an issue. Thanks again for the awesome tool, and please ping me if you decide to go after this. Cheers!

...also, just for a little more context, both issues produce patterns that impose a leading slash, or a double slash somewhere in the path. Here's an example produced from **/** as the raw pattern: https://regex101.com/r/KIDJtA/1

[cpburnz]

Hi @JonjonHays, thanks for the pull request.

For the first bug, there's a slight issue with it. It strips out all non-leading **s. E.g., **/a/**/b/** with segments ['**', 'a', '**', 'b', '**'] gets converted to ['**', 'a', 'b']. All that's needed to fix it is to add a or trimmed_segs[-1] != '**' to the if statement. I'll fix this. No need to modify the pull request.

For the second bug, that's an interesting case. Both the patterns ** and **/** have the regex '^.+$' yet **/ has '^(?:.+/)?/.*$' instead. I have an idea of how to handle this which I'll experiment with.

[cpburnz]

@JonjonHays I have a fix for the **/ pattern. You're test case was correct.

[JonjonHays]

Thanks for the update/merge @cpburnz ! Much appreciated =)