_read_deletes filters the full delete table once per distinct file_path

[sungwy]

_read_deletes builds its result with a dict comprehension that filters the whole table once per distinct file_path:

{
    file.as_py(): table.filter(pc.field("file_path") == file).column("pos")
    for file in table.column("file_path").chunks[0].dictionary
}

Cost is therefore O(rows x distinct paths). A delete file with many distinct paths makes this quadratic. The ORC branch below has the same shape via unique().

---

Issue investigation generated via claude, reviewed by Sung, Kevin, Fokko.