Implement a safe_url based on all standards

[Gallaecio]

Continuation of #201.

Fixes #193.

Tasks:

• Provide an alternative implementation for safe_url_string based on the URL living standard
• Add missing typing information
• Provide comprehensive test coverage for both implementations, highlighting the differences, and in the process make the new implementation support also RFC 2396 (used by java.net.URI) and RFC 3986 (the previous standard, still popular).
• Ensure the following issues are addressed by the new implementation:
  • Pipe symbol ("|") is not percent encoded #80
  • safe_url_string handling IPv6 URLs #193
• Have the URL living standard parse implementation mostly pass the upstream tests.
• Improve performance
  The new implementation is currently around 3-4 times slower than the previous implementation for test cases where both have the same, non-error outcome.
• Address to-dos in test_parse_url.
• Clean up the implementation: cleaner code, docstrings…
• Discuss what to do API-wise (deprecate safe_url_string in favor of safe_url? Remove safe_url and make safe_url_string use the new implementation?

[Gallaecio]

I am having a hard time lowering the gap below the current one, which is 4-3 times slower than the previous implementation 😞

[Gallaecio]

Looking at scrapy/scrapy#1306, maybe we could try going for https://chromium.googlesource.com/chromium/src/+/HEAD/url/ .

[Gallaecio]

Interesting read: https://daniel.haxx.se/blog/2022/01/10/dont-mix-url-parsers/

[kmike]

An interesting project: https://github.com/TkTech/can_ada

[albertedwardson]

can be useful: https://github.com/crate-py/url