[FEATURE] Add comment to identified license if match is not perfect

[jaimergp]

Is your feature request related to a problem? Please describe.
grayskull will identify licenses by fuzzy match in some cases and then suggest the best result. However, sometimes the fuzzy match scores a bit lower than average because some people modify popular licenses like BSD or MIT and add their own clauses. This kind of change will turn those into a different license altogether, which should be named as LicenseRef-packagename.

Describe the solution you'd like
I suggest:

• Adding a comment next to the suggested identifier, saying that its match % was lower than usual license: MIT # warning! license might have been modified; search match is XX%.
• If no clear license was found, generate a LicenseRef-packagename identifier with a comment next to it like # license could not be identified, please review.

Describe alternatives you've considered

Erroring out seems excessive.

Additional context

A recent example I found on a conda-forge submission.

[marcelotrevisani]

The match will never be perfect, so it will always need to add that comment

[marcelotrevisani]

Reopening that just because it still needs to add the comment
But @ForgottenProgramme already added a warning about it. :)

[bilderbuchi]

It seems to me that grayskull should not put too weak assumptions into the generated recipe.

In my case, I'm running grayskull over an sdist with a proprietary license, but grayskull does

Matching license file with database from Grayskull...
Match percentage of the license is 9%. Low match percentage could mean that the license was modified.
License type: BSD-2-Clause
License file: ['LICENSE.txt']

and puts

about:
license: BSD-2-Clause
license_file: LICENSE.txt

into the recipe.

While of course the LICENSE.txt should be the final (correct) truth, putting this down as a BSD-2-Clause in the package metadata is very far away from the truth.
I think the second bullet above about what to do when no clear license could be identified is very relevant, I'd expect a license: LicenseRef-LICENSE.txt or similar in my case.

[bilderbuchi]

Additionally, I've found no way to override the license detector when invoking grayskull, which could be a stopgap measure to avoid putting wrong metadata in the recipe. Is there one already that I missed?

[bilderbuchi]

Additionally, I've found no way to override the license detector when invoking grayskull, which could be a stopgap measure to avoid putting wrong metadata in the recipe. Is there one already that I missed?

For anybody finding this in the future, one workaround I found when using conda build afterwards, one can use this option to override the wrong license entry with an appropriate file, so at least the built package is correct:

  --clobber-file CLOBBER_SECTIONS_FILE
                        Clobber data in meta.yaml with fields from this file. Jinja2 is not done on clobbered fields.