Reject invalid URIs in URI extract functions

[findepi]

It's not by (any document) design to silently ignore invalid input in URI extracting functions (url_extract_fragment, url_extract_host, url_extract_parameter, url_extract_path, url_extract_port, url_extract_protocol, url_extract_query).

This commit follows least surprise principle: invalid input is explicitly rejected. Users wanting to ignore invalid input should use try(...) expression around the invocations.

This commit does not change docs, since now the functions behave as documented.

Release notes

(x) Release notes are required, with the following suggested text:

# General; potentially breaking change
* Fix URI extraction functions (``uri_extract_*``) not to ignore input which is not valid URI.
  Previously the functions undocumented behavior was to return ``NULL`` in such case.

[kokosing]

% comment

[kokosing]

test also try(url_extract_fragment(...

[findepi]

i don't think we should be testing try in TestUrlFunctions.java

[findepi]

CI #17436

[martint]

It's not by (any document) design to silently ignore ...

Actually, it was by design. It's not an accident that the implementations were doing return (uri == null) ? null : .... These functions were added before try() existed, so it was the only way to handle invalid data without the performance penalty of attempting to parse the URL first (e.g., via some hypothetical function like "isValidUrl()") before extracting the components.

This is a backward incompatible language change that deserves more careful consideration, and possibly a migration path that allows restoring the legacy behavior.