Simple version of the new GPU backend (#458)

* newgpu

* more to delete

* hmm, tests pass with constant folding

* fix lint/type

* fix constant folding

* comment and rerun tests

* lazy touchups

* fix graph_batchnorm test

* smaller transformer to fix OOM

* Revert "smaller transformer to fix OOM"

This reverts commit a44ef8edc2.

* no func cache

* introspect

* touchups

* CLASTKernel

* ugh, it was lru_cache

* codegen

* spacing

* old gpu still in opencl

* typing fix

test/graph_batchnorm.py

@@ -6,8 +6,8 @@ import unittest
 def model_step(lm):
   Tensor.training = True
   x = Tensor.ones(8,12,128,256, requires_grad=False)
-  loss = lm.forward(x).sum()
   optimizer = optim.SGD(get_parameters(lm), lr=0.001)
+  loss = lm.forward(x).sum()
   optimizer.zero_grad()
   loss.backward()
   del x,loss