Add deleteSet and difference operations

[edmundnoble]

These operations allow efficient deletes from an IxSet. These are often useful to filter an IxSet in a way that doesn't require rebuilding indices.

[adamgundry]

Thanks for proposing this! I'm also using ixset-typed in a way that would benefit from efficient (filtering and) deletion.

I notice that Data.IxSet.Typed.Ix.deleteSet is at least linear in the size of the map (i.e. the number of distinct keys), and Data.IxSet.typed.difference ignores the index maps. Couldn't we use the index maps to know which specific keys contain elements being deleted, and thereby have something more efficient, in particular in the case of large-ixset \\ small-ixset?

Also, it seems there isn't a filter operation yet, but presumably one add it along the lines of this PR?

cc @kosmikus

[edmundnoble]

I hadn't thought of that. That seems like a strict improvement. Thanks @adamgundry. I'll give that a go.

[edmundnoble]

@adamgundry I've done as you asked. I'm very glad to hear that someone else would benefit from these functions. I have also done a follow-up which adds filter.