Pyranges

- fix fisher exact when given pd.Series
- fisher_exact: only use pseudocounts for OR

- add outer, inner and left join
- add fisher exact
- insert series/dataframes to pyranges with + operator
- gr.Whatever = pd.Series(...) now ignores index
- add gr.copy() method to create deep copy

- add k-nearest
- ensure that start/end have the same dtype after calling slack
- breaking change: new_position takes no default arg
- new_position takes an argument swap
- .length returns a python integer
- breaking change: lengths returns a vector by default

- fix attributerrors on pyranges (thanks https://github.com/MuhammedHasan)
- add reader for gff3
- add writer for gff3
- add count flag to cluster/cluster_by

- fix merge print functions
- make pickleable
- add iter as alias for itergrs in pr. namespace
- gr.length() shows nucleotide length (sum of all interval lengths)
- gr.lengths() takes as_dict=False flag to return as vector
- fix slack in join: added columns when joining with itself
- fix print for unstranded pyranges: printed tail and head of first chromosome

- add overlap-flag to tile
- add chain to print-method
- bugfix: printing stranded pyranges sorted output even though sort was false
- bugfix: wrong number hidden cols on very small terminal widths
- bugfix: unstrand did not change underlying dict to chromosomes only
- show number of hidden columns in header
- tests: mismatches in strand between dict and dataframes
- .df/.as_df() now returns with non-duplicated index