Release 1.0.2 with the drop-empty fix

[clayote]

Though I was able to work around bug #55 in 1.0.1 like this:

import pyarrow as pa
from parquetdb import ParquetDB
from pyarrow import compute as pc

db = ParquetDB('parquetdb')
if db.is_empty():
    return
unwanted = pa.table([[]], schema=pa.schema([("id", pa.int64())]))
for datum in data:
    exprs: list[pc.Expression] = []
    for k, v in datum.items():
        expr = pc.field(k) == pc.scalar(v)
        exprs.append(expr)
    selected = db.read(filters=exprs, columns=["id"])
    unwanted = pa.concat_tables([unwanted, selected])
if unwanted.num_rows == db.n_rows:
    db.drop_dataset()
else:
    db.delete(ids=unwanted["id"])

This does seem like an unnecessary piece of boilerplate, and might affect performance in apps that delete data often.

So I've made a fork of ParquetDB that applies the real fix, #55, and only that fix. This PR points to that fork, but I don't think you want to merge it, as such, since you've done significant refactoring after 1.0.1.

Please clone my branch 1.0.2 and do a release.

Fixes #58.