deep copy in make_browser_figure

[thekugelmeister]

Avoid modifying the input dataframe in make_browser_figure, to avoid a bug where it was impossible to rerun the browser figure with collapsing reads until the input dataframe y_index column was reset to be non-unique.

The current (very simple) implementation of this fix relies on a deep copy operation, meaning that for very large datasets this may cause memory issues. I do not foresee this being the bottleneck for read browsing; by the time this becomes an issue the browser itself may be completely inoperable. However, I am leaving the space for a different fix to be implemented if this is a problem.

Addressing this issue in some other way will require a rewrite of the plotting logic, as the column in question is referenced inside of a different groupby operation.

[OberonDixon]

This seems like a fine approach to me. The change I am requesting is to get rid of or make non-redundant the .copy() operation, which I believe currently is creating a copy that does not then get used.

I assume that the normal level of copying depth that gets applied by .assign is ok, because we aren't making in-place modifications to python objects (such as numpy arrays) downstream in the pipeline. If I am misremembering though then it could be that we need a deeper copying operation.