Speed up DictionaryBrowser

[francisco-dlp]

As discussed in #363, DictionaryBrowser is currently the performance bottleneck for Signal creation and, therefore, for indexing.

One way to speed it up is to make it lazy, i.e. the DB machinery is not run until the __getattribute__ method is executed.

[to266]

Here I found a similar thing here
It seems to be a bit quicker to create as well:

    In [1]: from hyperspy.misc.utils import  DictionaryTreeBrowser
    In [2]: import attrdict
    In [3]: life = {'bigBang' :{'stars':{'planets': {}}}}
    In [4]: %timeit a = attrdict.AttrDict(life)                                                                   
    10000 loops, best of 3: 54.6 us per loop
    In [5]: %timeit b = DictionaryTreeBrowser(life)
    10000 loops, best of 3: 89.1 us per loop

obviously it does not have all the same features (i.e. printing, set_item, etc), but that can be possibly added if needed.
From a quick glance, the implementation seemed to be different, though, which is a good thing

[jlaehne]

@ericpre can we close this issue now?