Refactor haskell-navigate-imports

It should work like this:
- first M-g i pushes mark using `push-mark` and goes to first `import` in file
- repeated M-g i moves to the next import block (group of import statements separated by something else)
- repeated M-g i cycles to the first one when past the last one
- standard C-u C-SPC is used to return to old point position

Half of the logic should go away.