Consider include a source-free or otherwise-stripped XML tree with get_source_expressions()

[MichaelChirico]

As a general thought for the future: would it be terrible for performance to add a new xml_parsed_content_nocomments elements in the output of get_source_expression()? To avoid the gymnastics in each individual linter.

Originally posted by @Bisaloo in #2901 (review)

Having to remember whether to strip out comments is a pain. What's the right way to fix this in get_souce_expressions()? My memory is keep.source=TRUE is required for getParseData() to work, and anyway fully stripping out all line+column info in addition to COMMENT nodes is probably a step too far for the task of skipping over comments.

[Bisaloo]

What about running strip_comments_from_subtree() on xml_parse_content in maybe_append_expression_xml()?

It may cost some time to run this on all expressions, but at the same time, we would run only it once, and not on a possibly redundant subset of expressions for each linter, as we are doing now.

Any reason why this wouldn't work?