Ast opencl fixes

[sakehl]

Mainly added support for rewriting nested foralls, which are automatically generated by GPU kernels and par blocks.
This is done in the pass SimplifyNestedQuantifiers.scala.

This pass relies on ExpressionEqualityCheck.scala, to compare expressions on equality, whilst getting information like require blockSize == 32 from annotations.

There are some changes in how a GPU kernel is encoded (LangCToCol.scala) and how par blocks are encoded (ParBlockEncoder.scala), to make sure that par blocks and GPU kernels require to have at least one thread running (therefor, we don't have to account for "void" runs of GPU kernels). Which was needed to gather information, which is needed in the expressions equality check, which was used in the simplified nested foralls.

Missing are some examples and unit tests still.

Also a lot of other minor changes are present:

• added bool type to opencl and cuda headerse
• After quantifiers are rewritten in simplify relations, to original name is kept (or at least we try to)
• C arrays and variables will keep their internal name
• Some things I probably forgot I changed.

[pieter-bos]

It's taking me a bit longer than expected to review everything, so here is a partial review so it doesn't come late in one batch :)

[sakehl]

It's taking me a bit longer than expected to review everything, so here is a partial review so it doesn't come late in one batch :)

I fixed the changes you suggested. Only 2 things remain open, see above. Also added barriers with memory fences