TODO: feature/sequences branch

I'm just going to start adding notes/nits here as I look over the sequences code. Things to do / clean up / discuss / etc. before the merge into master.

- [ ] There are so. many. warnings.
- [ ] Stylistic changes; e.g.
- spacing and alignments (consistency)
- layout / organisation; e.g. helper functions defined far from their use site (for no reason)
- [ ] The tuple fusion story
- `eltFlavour`; yuck.
- The comment in AccelerateHS/accelerate@11622cb7758872c756f2ba33ed2770403c86609e is wrong, so I'm not sure what the motivating use case for disabling this was (I'm sure there was one, I want to know what it was).
- [ ] Check for other "temporarily disabled" things which should be addressed.
- [x] `SeqIndex` offends me (tmcdonell/accelerate@a2efe9e706337854c239f0ebc15ea0ebdf5f8858)
- [ ] graphviz support
- [ ] `duplicateToRemote` seems sketchy
- [ ] `useLocal`: are sequence computations copying back _intermediate_ computations to the host?? ([here](https://github.com/AccelerateHS/accelerate-llvm/blob/54057bc3590bdcfc2513ee4c79e92446e8a64175/accelerate-llvm/Data/Array/Accelerate/LLVM/Execute.hs#L600))
- [ ] lots of missing documentation, both for internal stuff as well as exposed (surface language) functions
- [ ] Comment what [this](https://github.com/AccelerateHS/accelerate/blob/4c74a80a836bce82dcbcbe79ad6d1fa1a3e5fa24/Data/Array/Accelerate/AST.hs#L555) is
- [ ] nofib/sequences module is [poorly laid out](https://github.com/AccelerateHS/accelerate-examples/blob/56b2ad1afd1e0923477ddac92b871bf76a031f2d/examples/nofib/Test/Prelude/Sequences.hs#L268)
- [x] use of `chunk-size` in debug flags is extremely suspicious. Also its implementation is suspicious; I know that it is currently the only flag with `Maybe Int` type, but needing to special-case this in the parser is clearly the wrong way to do it. (tmcdonell/accelerate@17511264ec050a7a8ba98fc2a33385337561bab0)
- [ ] replace `IndexTrans` with `IndexLast`, `IndexSnoc`, `IndexInit` (?)