Functions are included under `mungesumstats.jl`. Consists of two user
facing functions: `readchain` to read a chain file for liftover, and
`liftover_sumstats!` to liftover a munged summary statistics DataFrame.
To do includes handling the edge case where the target strand in the
chain file has a negative strand and adding examples into the
documentation.

`liftover_gwas!` is now `liftover_sumstats` to be in line with
`mungesumstats!`. Additionally, `liftover_gwas!` can now be applied to
AbstractVector{<:AbstractDataFrame}, in line with `mungesumstats`
behavior. `parsechain` drops the "chr" prefix to be in line with
`mungesumstats!` behavior. Function parameter `echain` renamed to
`chain` (originally named `echain` for "expanded chain", but was
needlessly verbose since we never use the output from `parsechain`).
Documentation for the liftover functions added to the end of the summary
statistics tutorial and mentioned in the GENCODE GTF parsing tutorial.
Tests updated to test liftover functionality over a vector.

All the other functions omit underscores; rename to match style.

Rewrote `findnewcoords` to sort and then iterate through the summary
statistics in order to liftover more efficiently. `readchain` also
sorts the chain file now for that same purpose. This replaces the
slower old code which used DataFrames.subset to find matching regions
between builds. Also, added a bug fix to `readchain` where the ending
coordinate was included: the chain format uses half-open intervals.

Moved liftover code from mungesumstats.jl to liftover.jl. Added some
more error catching code (mainly for when lifted over indels no longer
match the reference).