Adds performance annotations using the new Hatchet annotation interface

[ilumsden]

This PR adds performance measurement annotations to Thicket using the APIs added to Hatchet in LLNL/hatchet#142.

[dyokelson]

@ilumsden Can you remind me - is there anything important about the order of the decorators or are they applied independently? Just wondering what the outcome will be for some functions with the cache_stats_op and perf annotations.

Otherwise looks good, since this is just annotations and the functionality is in hatchet.

[michaelmckinsey1]

A couple of comments:

1. Is it possible to add metadata with the pycaliper interface?
2. I think the amount of annotations is causing high overhead. See following:

Time to create Thicket with X amount of files.

files	pycaliper	develop
1,000	36s	8s
10,000	15m36s	1m26s

I notice Calls for Node.__eq__ is 96,174,300 for the thicket with 10,000 files. By removing all of the node annotations in Hatchet for pycaliper time goes from 15m36s to 3m. So I think we may have to be more selective with what we annotate or maybe provide some kind of different levels for profiling.

[ilumsden]

Thanks for taking a look @michaelmckinsey1. That timing info is very interesting. I definitely agree that the annotations need to be paired down given the timing.