Make the file limit configurable

[pkkm]

Currently, Linguist has a 100,000 file limit hardcoded in lib/linguist/repository.rb:

    protected
    MAX_TREE_SIZE = 100_000

When the number of files in the repository exceeds that, Linguist gives up and returns an empty result:

    def compute_stats(old_commit_oid, cache = nil)
      return {} if current_tree.count_recursive(MAX_TREE_SIZE) >= MAX_TREE_SIZE

This may be a good limit for GitHub, but I think that the library would be more widely useful if it was possible to change it—for example, via a getter/setter pair that accepted either a number or nil (to disable the check entirely).

A related issue is that when the limit is exceeded, Linguist returns an empty hash, which is the same as for an empty repository, with no indication of what happened. This can be quite confusing—or at least, it confused me the first time it happened, and issues #2915 and #3342 indicate that it has confused other people as well. Perhaps it could be changed to something that communicates the issue clearly, such as return :too_many_files or raise FileLimitExceededError?

[lildude]

Resolved by #6806 and #6825