Implement bubble finding in FlatGFA

Let's add functionality to FlatGFA to find _bubbles_ and _superbubbles_. The end goal is to create a drop-in replacement for [the BubbleGun library][bg], or at least its `find_bubbles`, `connect_bubbles`, and `find_parents` functions.

As a brief reminder, a "bubble" is a pattern in a variation graph where there are multiple distinct paths between two vertices (segments). For more background on how bubbles are defined, see [the BubbleGun paper][bg paper] and [the flubble paper][flubble paper].

* [ ] Gather some background:
* Read the two papers above, including [the supplemental material for the BubbleGun paper][bg sup]'s Algorithm 1, to understand the definitions and algorithms involved.
* Download and try out [the BubbleGun tool][bg] so you understand how its various commands work.
* [ ] Implement the BubbleGun algorithm. Add a subcommand `fgfa bubbles` that uses the graph-traversal algorithm specified in the supplemental material.
* [ ] Add _differential testing_. The idea is to check that BubbleGun and `fgfa bubbles` produce the same answers on a few example graphs.
* [ ] Measure performance. Using a collection of larger graphs, compare the performance of BubbleGun with our implementation. Hopefully, we'll be faster! If not, we'll need to do some optimization at this stage.
* [ ] Add Python bindings. We would like to be able to call the above functionality from Python so that we can provide more of a drop-in replacement for the BubbleGun Python library.
* [ ] Explore the alternative definitions and algorithms in [the flubble paper][flubble paper] and [its implementation in the povu tool][povu]. Consider implementing these alternatives and comparing (correctness and performance) against povu.

At some point in this process, we will also approach the authors of [PangyPlot][] to see if our drop-in replacement could help them.

[bg]: https://github.com/fawaz-dabbaghieh/bubble_gun
[bg paper]: https://academic.oup.com/bioinformatics/article/38/17/4217/6633304
[flubble paper]: https://arxiv.org/abs/2410.20932
[bg sup]: https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/bioinformatics/38/17/10.1093_bioinformatics_btac448/2/btac448_supplementary_data.pdf?Expires=1763701898&Signature=cW0PPb2cIaSp7FIQ3oLJxL9curPdXXM~oyq2QJ2IlnCEOd~G747s6ZwK3-SwDhLuiGBt-W-XYTHS4Ug4v2GZNJ6ZLsapagN9twfsxabAfP0uJdKGG5WY9VjZhCZExd6Do-XKWlY9PbP3HkwP-ve1JL~HcI5QrlI3Ccf2qnZLZ3-E4SCBG19KTGy~ESQvC4fn7y-ih00VUUyZdQ5eajbw23rCSC9yRTkiezoImxxqFIr7WY77Ve-donAT~0YcXpr0wbfwxJeysRasDeWk82t-vH-c5MaJtW0mpYLw4rhGeEQy1L~8pFm2BUVYLJr6NuC8nKqDc1Nu-Jcpwl1Q7iJIaw__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA
[povu]: https://github.com/urbanslug/povu/
[pangyplot]: https://github.com/ScottMastro/pangyplot

Start by reading the BubbleGun paper and supplemental Algorithm 1, then try the BubbleGun tool to understand find_bubbles, connect_bubbles, and find_parents. The issue names no repository files or tests; a complete effort would define the fgfa bubbles subcommand, differential tests, performance comparisons, Python bindings, and any selected flubble or povu alternatives.

Written by the indexing model from the issue text.