BF: Load gifti parser iff parsing gifti files

[effigies]

Setting parser as a class attribute turns out not to be very necessary. Avoid the circular reference by calling the parser directly when needed.

Fixes #392.

[effigies]

Couldn't figure out a new test to write, since loading files is already tested. Could del GiftiImage.parser before trying to load, but this change removes the attribute altogether, so the test would be obsolete as soon as it's written.

[bcipolli]

The intention here was to set a class-level property, similar to how header_klass is set on images. So now the corresponding parser class isn't easily discoverable for an image requiring one.

With that said, LGTM! This fixes a bug, and I don't see that having the parser class set explicitly on the class will matter anytime soon. Thanks @effigies !

[effigies]

Definitely open to alternatives, but my only other idea was to just merge the two files together and eliminate the import problem.

I could add a comment, in case somebody decides to come back to this, if that would be helpful.

[bcipolli]

Right. It's confusing. My solution wasn't the best either.

I think what you've done, with an informative comment, is the best choice available.

[bcipolli]

("it's confusing" === "the best solution is hard to figure out")

[matthew-brett]

Ouch - nasty - the Image class needs a Parser class and the Parser class needs the Image class.

I stared at it for a while - what do y'all think of this alternative?

master...matthew-brett:alternative-gifti-fix

[effigies]

@matthew-brett 👍 Looks like a good alternative. No suggestions from my end. Feel free to close this PR if you go ahead with it.

[matthew-brett]

Closed by #406