Implement IGD

[nleroy917]

We need to re-implement IGD in this crate. Being done by @donaldcampbelljr in #9

Original code here: https://github.com/databio/IGD

[nleroy917]

For python bindings, we could do an OOP approach:

from gtars.igd import Igd

igd = Igd.create_from_files(
    source_files="path/to/files",
    output_folder="path/to/output",
    database_name="mydb"
)

# way later
igd = Igd.load_db("path/to/database)
idg.search(...)

[donaldcampbelljr]

IGD create and search now work in PR #9 with some caveats.

An IGD database can be created from a folder full of bedfiles. A search can be performed using a single bed file as the query.

Performance-wise, creation appears to be similar for C and Rust versions (80 files, ~280,000 regions) at 2.1 seconds.

There are some discrepancies between the C version that should be investigated in the future such as:

• during searching, the number of counts may be off for one or more of the bedfiles. Most are the same, however. This may be due to pulling the .igd data back into memory for query. The gData tiles are not 1 to 1 (see attached picture). The creation step appears to be exactly the same based on numbers (e.g. # of Ctgs, Regions, Tiles, etc).
• Certain bedfiles may cause the C version to crash but will pass using the Rust version and vice versa. Anecdotally, the Rust version appears to be more robust.
• Test performance for larger sets of bedfiles.

[donaldcampbelljr]

I just merged the PR that has been in progress since beginning of the year. However, IGD still needs some work.

• feature parity with C version
• understand any discrepancies between this version and C version -> See: IGD Search is returning 0 hits in some test cases #45
• python bindings were not yet implemented

[donaldcampbelljr]

Closing this in favor of more specific issues for IGD work.