Add script to simplify adding a missing type.

[dosas]

Should simplify: #45

[h2non]

I'm not very convinced about this code generation approach.

It will lead to inconsistency in the way to declare matches in the library unless all matches are ported to be defined the same way consistently within the package, which is not something I'm very inclined to. Typically, dynamic code generation has its advantages where relevant boilerplate steps are involved, but I don't see much value here. I would rather prefer the explicitness of persisted and evaluable code without any code generation process involved.

[dosas]

I get your point about the inconsistencies. You mean the comparison by bytearray versus the or, and clauses?

I disagree on the boilerplate because in order to add 4 bytes, one extension, and a mime type I have to write about 8 lines of code (that only differs in the mentioned properties), add the type to the README, and to __init__.

I don't get this point:

I would rather prefer the explicitness of persisted and evaluable code without any code generation process involved.

The generated code is committed (I did all my PRs using this generation appraoch)