Future versions of the Markdown processor don't include per-line indent
information. Thankfully, we're already calculating it in
`getBlockRangeMap()` because we need it to correctly map fix ranges for
lines that are under-indented to the left of their opening code fence.
This includes the indent in the `RangeMap` for each line and uses that
in `adjustBlock()` instead.

This uncovered an incorrect assertion in one of the tests, which
combined an indented code block, a message on a line with mixed spaces
and tabs indentation, and an under-indented final line. The message
column was incorrectly asserted as being in the whitespace indentation
before the line.

Originally discovered in #120 and cherry-picked from #125.

The previous parser, `remark-parse` v7, included a transitive dependency
on an npm package with a security vulnerability. Newer versions of
`remark-parse` are wrappers around a new underlying parser,
`mdast-util-from-markdown`, so we can use that directly.

The previous parser also failed to preserve `\r\n` line endings,
replacing them all with `\n`. The new parser correctly preserves `\r\n`
line endings, finally providing a fix for the failing test case I
cherry-picked in the previous commit. The improved behavior also
uncovered an incorrect line ending test assertion that this commit
corrects.

While this change is in theory fully compatible, containing just bug
fixes, I'm tagging it `Update:` in case there are compatibility changes
in the new parser. This is consistent with #175, which upgraded
`remark-parse` v5 to v7 in a semver-minor `Update:` change.