specifying distance functions

There is an inconsistency with how distance functions for different feature dimensions are handled. When an `egg` is created, the user can optionally pass a `dist_funcs` dictionary containing distance functions for each feature. This is used in the clustering analyses (fingerprint, temporal). However, when performing naturalistic analyses, a new distance function is assigned, e.g. `egg.analyze('spc', match='best', distance='correlation')`. This forces the distance function to be 'correlation' for all features. Some possible resolutions:

1) keep the distance argument, but default to `egg.dist_funcs`. if a string or function is passed, assume that all features should use that distance function. if a dictionary of feature:function pairs is passed, use these instead of the functions specified in `egg.dist_funcs`

2) get rid of the distance argument in the `analyze` method (always infer it from the egg.dist_funcs dict).

1 seems like a more flexible solution...by default assumes you want to use `egg.dist_funcs`, but allow the user to modify the distance functions on the fly. @jeremymanning and @paxtonfitzpatrick , can I get your thoughts on this?