Verse segmentation

[benjaminking]

This PR adds a new script called segment_verses.py that can segment passages into individual verses by aligning the passages against known verse segmentations (i.e. any Paratext project). It is also able to evaluate the accuracy of the segmentation.

---

This change is 

[Enkidu93]

Still reviewing the algorithm, but I can continue to do that once this is in :). Looks great - just a couple small comments.

@Enkidu93 reviewed 2 of 2 files at r1, all commit messages.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on @benjaminking)

---

silnlp/alignment/group_verses_into_passages.py line 71 at r1 (raw file):

def main() -> None:
    parser = argparse.ArgumentParser(description="Collect verse counts and compute alignment scores")

Is this description correct?

---

silnlp/alignment/group_verses_into_passages.py line 74 at r1 (raw file):

    parser.add_argument("--project", help="Name of source Paratext project", required=True, type=str)
    parser.add_argument("--input-passages", help=".tsv file with target passages", required=True, type=str)
    parser.add_argument("--output-passages", help=".tsv file with target passages", required=True, type=str)

Should this say 'source passages'? I might not be following correctly.

---

silnlp/alignment/segment_verses.py line 85 at r1 (raw file):

        for line in load_corpus(passage_file):
            row = line.split("\t")
            if len(row) < 7:

Maybe != for robustness - not sure it matters much

---

silnlp/alignment/segment_verses.py line 247 at r1 (raw file):

                continue

            fewest_crossed_alignments = 1000000

Not necessary, but a cool thing you can do that you might not know about is use underscores in numbers - e.g., 1_000_000 - doesn't change the value, but makes it easier to interpret at a glance.