That's the first step. The modified files are determined via pull/{pull_number}/files.
This API call returns at most 3000 files. We use a page size of 30 and collect at most
10 pages, so that we have a total of 300 files - that's what we use for analyzing.
This means for big changes, not all files might be analyzed.

Also add logging which mode is active.

This also sets the output parameter to zero.

Usually debug logging is enabled by adding a repository secret
"ACTIONS_STEP_DEBUG" with value "true". However, this is
secret not available if a PR runs from a forked repository...

This reverts commit 7f99864.

Let the unit tests run under macos and windows additionally.

The modified files are determined via compare API call. Similar
to pull requests, we only collect at most 10 pages. However, the
number of total changed files can't be calculated straightforward,
since the pagination happens on the basis of commits rather than files.
In any case, for big pushes this means, that not all files might
be analyzed.

This always logs a warning if we fetched page MAX_PAGE
regardless whether there would be more pages to fetch
or not. No effort is done to determine
whether the list of modified files is complete.

Also the modified files are now collected in
a Set to avoid duplicate entries.