Try to implement copy-paste protection checks

[Martoon-00]

Clarification and motivation

Imagine the following list of links:

• [file](files/file.out)
• [file2](files/file2.out)
• [file3](files/file3.out)
• [Another file](files/another-file.out)
• [And a file once again](files/and-a-file-once-again.out)

It is easy to make a mistake here during copy-pasting so that text is updated and the link is not. I think we can use some heuristics to spot such mistakes (but avoid false positives at all costs):

• If this check is enabled in config (it should be by default);
• And there are two links [T1](L1) and [T2](L1) in a file, and T1 is substring of L1 modulo casing and all the non-alphanum characters, while T2 is not substring of L1 modulo the same things;

then report an error at [T2](L1) position, mentioning that it could be a bad copy-paste of [T1](L1). And a similar check for [T1](L2).

Acceptance criteria

• The check is implemented.
• It can be disabled both in config and in place via some <--! xrefcheck: no duplication check in {file/paragraph/link} -->.

[Martoon-00]

If possible, we should check reference-style links ([text][link-id]) too.

However not sure how much is it possible, AFAIR such links are automatically inlined by our markdown parser.

[YuriRomanowski]

Do we want to check only links within one list? How about checking all the links within a given file?

[Martoon-00]

That's a good question. On the one hand, this increases the probability of getting a false positive. On the other, checking through the entire file may be more useful and will be a more transparent behaviour for the user.

Let's really go with checking across the entire file.

Over time we will collect some statistics on how this check works on real-life repositories and will revise the behaviour then.