Pyranges

- fix regression: slack changes dtype from int32 to int64

- add dtypes attribute to pyranges
- fix left and right join when chromosomes missing

- add argument sparse to read_bam. Setting it to False fetches more columns.

- fix column names after read_gtf so they work with GenomicFeatures
- add flag chain, make False by default to to_* methods
- genomicfeatures: add tss/tes-methods
- fix column names after read_gtf so they work with GenomicFeatures
- remove Strand column with unstrand() even if PyRanges is not stranded
- reading gff and gtf now consistent and column names from attributes are in lower_snake_case

- add missing example data (ending with gz) to pyranges
- add rowbased_spearman, rowbased_pearson and rowbased_rankdata to pyranges.stats
- pyranges now accept columns with integer names, like pandas

- ignore index when inserting Series
- able to add dictionary of dfs to a pyranges
- remove FDR from fisher_exact, but add fdr as own method in stats
- make stats.mcc faster
- make stats.mcc work without a genome
- fix gff3 reading when metadata contains spaces