Fix existing percent encoding

[samuelcolvin]

Change Summary

As described in the issue, #4458 was double encoding URLs, this splits the URL on encoding groups with the regex (.*?)(%[\dA-F]{2}) to (hopefully) fix the issue.

Changes:

• existing "percentage encodings" are not re-encoded - they're just cast to upper case to match RFC 3986 and httpe - e.g. foo%2Bbar is left unchanged while foo%2bbar is converted to foo%2Bbar
• forward slash / is no longer encoded in query paramters, e.g. http://example.com?a=b/c is left unchanged

Related issue number

fix #4458

Checklist

• Unit tests for the changes exist
• Tests pass on CI and coverage remains at 100%
• Documentation reflects the changes where applicable
• changes/<pull request or issue id>-<github username>.md file added describing change
• My PR is ready to review, please add a comment including the phrase "please review" to assign reviewers

[hramezani]

I think we still will have the problem reported in #4468 even with this patch. which is a breaking change.

[samuelcolvin]

Okay, this problem is bigger than I thought due to the strange logic to decide when to rebuild urls, see

pydantic/pydantic/networks.py

Lines 297 to 300 in 317bef3

	host, tld, host_type, rebuild = cls.validate_host(parts)
	return cls(
	None if rebuild else url,

and

pydantic/pydantic/networks.py

Lines 377 to 379 in 317bef3

	if is_international:
	host_type = 'int_domain'
	rebuild = True

which is completely out of date from the URL quoting - we previously only rebuild domains when required to cope with "international" TLDs, but now we should be rebuilding if there's a path or a query or a user or anything which might need quoting.

However if I just force rebuilding of all URLs, lots of tests are failing.

[samuelcolvin]

I think we still will have the problem reported in #4468 even with this patch. which is a breaking change.

Good point, should be fixed by the last commit.

[samuelcolvin]

replaced by #4470.