Pyranges

Hotfix:
- left in dbg statements

Hotfix:
- unstrand() did not always remove strand info

Additions:
- pr.PyRanges() returns empty PyRange before you needed pr.PyRanges({})
- pyranges are now callable. Examples: gr("df.Score > 0") and gr("df.A.astype(str) + mysuffix")
- can subset PyRanges with a dict of boolean vectors
- pr.data.exons(), pr.data.cpg()
- gr.unstrand() removes strand information from a PyRanges
- throw exception if trying to drop Strand from df without setting drop_strand=True
- adding a Strand column to the PyRanges makes it stranded

Changes:
- write dtype as category, not int8/int16/...

Fixes:
- remove empty dfs in the dict given to the PyRanges constructor

Removed:
- gr.data.epigenome_roadmap()

Additions:
- gr.cluster(): assign ID to each cluster found by merge
- gr.columns: return the columns in the pyranges
- gr.drop: drop columns based on regex or list
- gr[["Score", "Name"]]: select subset of columns
Fixes:
- gr.stranded errored if chromosomes were ints