Fix regex of URL quadratic backtracking

[Fazel94]

URL regex quardratic backtracking

Description

In spacy/lang/tokenizer_exceptions.py:16 the regex for mailto:user or user:pass auth faces takes quardratic time to match crafted examples like "a:"*16000.

Code to reproduce the issue:

import time
import spacy

nlp = spacy.blank("en")

for n in (1000, 2000, 4000, 8000, 16000):
    text = "a:" * n
    start = time.perf_counter()
    nlp(text)
    print(f"{len(text):>6} chars  {time.perf_counter() - start:7.2f}s")

On the current master would write:

  2000 chars     0.03s
  4000 chars     0.08s
  8000 chars     0.36s
 16000 chars     1.20s
 32000 chars     4.76s

On the PR it would produce:

  2000 chars     0.01s
  4000 chars     0.01s
  8000 chars     0.03s
 16000 chars     0.05s
 32000 chars     0.10s

For tests I have added two pathological cases and measure their time to prevent regression.

Checklist

• I confirm that I have the right to submit this contribution under the project's MIT license.
• I ran the tests, and all new and existing tests passed.
• My changes don't require a change to the documentation, or if they do, I've added all required information.