Remove dependency on itertools

[CosmicHorrorDev]

This PR removes the dependency on itertools in favor of slightly more verbose implementations. The two usages of itertools were for

izip!()

Which was just swapped for chaining .zip() followed by a final .map() to pack the elements into a single tuple

And .group_by()

Which was replaced with grouping everything into a BTreeMap first, then iterating over the key -> group pairs.

This implementation is slightly different since .group_by() will chunk the groups which means that the same key can appear multiple times if it is separated by another key. The implementation in this PR groups all values with the same key regardless of if they are separated by another key, so I'd like for someone else to double check that this still follows intended behavior

[lemmih]

Anyone available to double check that this still works as expected?

[briansmith]

I ended up submitting my own PR to do this, slightly differently. PTAL if you are interested: #876.

[CosmicHorrorDev]

Super happy to have someone else pick up the torch. Thanks @briansmith !