fix: measure MD038's code span against its own delimiter

[akiomik]

MD038 compared code.literal against a content length derived from the span's
columns:

let content_len = position.end.column - position.start.column - 1;

That arithmetic assumes a one-backtick delimiter. For a span delimited by N
backticks it overshoots by 2N - 2, so code.literal.len() != content_len can
never hold and every such span was reported. There is no way to author around
it: two backticks are how CommonMark writes a span that contains one.

What this changes

Measure the span rather than slice the line. sourcepos covers the delimiters
and comrak carries num_backticks, so the content is 2 * num_backticks bytes
narrower than the span; against literal that width is either equal — nothing
was removed — or two bytes longer, which is the one space CommonMark takes off
each end.

A width is used rather than the line itself because sourcepos is not always an
index into it. comrak unescapes a table cell before parsing its inlines, so
inside one every \| earlier in the cell shifts the columns after it by a byte.
A width survives that, because both ends shift together.

That leaves what counts as padding. CommonMark removes one space from each end
of a span that both begins and ends with one, and that pair is the only way to
write a span that starts or ends with a backtick:

`` ` ``

So a removed pair is reported only when it shields something other than a
backtick, and any space the removal leaves behind — `` wide `` — is
reported as before.

Behavior

Every line of the report from #391, before and after:

Input	Expected	Before	After
`ok`	clean	clean	clean
``ok``	clean	MD038	clean
``a `b` c``	clean	MD038	clean
```lit```	clean	MD038	clean
` pad `	MD038	MD038	MD038

A span whose delimiters sit on different lines is not judged

Nothing recorded about one describes what was written against its delimiters.
literal has already turned the line endings into spaces and taken the padding
off, and a width cannot be had from columns that index two different lines.
Reading the lines back was tried and fails four ways, each found in a real
document:

• the columns shift once a continuation line has been stripped of its
  indentation, and point past the closing delimiter
• the lines between the delimiters are not part of any slice of the two ends
• a container's > or list marker lands inside the slice and stands in for the
  line ending that is really there
• a span whose ends are both empty rebuilds as a single space, which reads as
  padding on a span whose spaces are the only way to write it

So the span is skipped. The cost is a miss on padding written across a line
ending; the alternative is reporting spans no author can act on, which is what
this pull request is about.

Verification

• cargo test --all-features --workspace, cargo clippy --all-targets --all-features --workspace -- -D warnings and cargo fmt --all --check all
  pass. src/rule/md038.rs keeps 100% line coverage. The two tests this rule
  already had are unchanged; twelve cases were added for multiple backticks,
  backtick content, a table cell after \|, a multibyte prefix, a span of
  nothing but a space, and each of the four multi-line shapes above.
• mado check . on this repository still passes.
• On the GitLab corpus under scripts/benchmarks/data — 10k+ files — MD038 goes
  from 56 reports to none. Every one of the 56 was a span with no padding in it.
• Against the acceptance fixtures under scripts/acceptance/data, MD038 reports
  the same three lines of spaces_inside_codespan_elements.md as it did before.

Closes #391.

🤖 Generated with Claude Code

https://claude.ai/code/session_01JWNoFSfDhBzXEX9UL68oqf

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.79%. Comparing base (63deb47) to head (03c2bc6).

Additional details and impacted files

@@           Coverage Diff            @@
##             main     #402    +/-   ##
========================================
  Coverage   99.78%   99.79%            
========================================
  Files          72       72            
  Lines        7485     7715   +230     
========================================
+ Hits         7469     7699   +230     
  Misses         16       16            

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.