Added font lock tests for pattern synonyms (for #1313)

[taktoa]

I wrote tests for the font lock issues with the pattern keyword, as discussed in #1313.

Please check if the way I used check-properties is correct, since:

• It wasn't entirely clear but it seemed like the triplets in the second argument of check-properties each describe the fact that a token should be fontified a particular way in all the test strings given in the first argument (the alternative interpretation is that the triplets describe a series of fontifications for each token in order, and since none of the existing tests duplicate the first element of a triple, it's ambiguous which it is).
• I gave t as the second element of each triple, which I am not sure about either. This element is the "syntax" of the token, but I don't know enough about Emacs fontification to know what that means.

The haskell-no-pattern-1 and haskell-no-pattern-2 tests currently pass, so I commented out the :expected-result :failed so as to prevent an error.

cc: @gracjan

EDIT: fun fact, this is pull request number 1337 on haskell-mode :D

[gracjan]

This checks only this first pattern. You need to write as many patter rules as there are pattern keywords in the example code. It is okay to leave out other tokens, check-properties skips over fragments that aren't listed.

[gracjan]

Thanks.

[gracjan]

Looks like I was wrong and when PatternSynonyms are enabled then pattern is a keyword everywhere it appears. For example this is a syntax error:

foo :: (a -> pattern) -> a -> pattern