tensor: raise error when setting requires_grad=True on non-float tensors

[kozuedoingregression]

fixes issue #9723

This PR adds a runtime check in the Tensor constructor to raise a RuntimeError when requires_grad=True is set on a non-floating point tensor.

from tinygrad import Tensor
print(Tensor([1.0, 2.0, 3.0], requires_grad=True).tolist())  
print(Tensor([1, 2, 3], requires_grad=False).tolist())       
print(Tensor([1, 2, 3], requires_grad=True).tolist())        

out:

[1.0, 2.0, 3.0]
[1, 2, 3]
Traceback (most recent call last):
  File "/mnt/z/work/tinygrad/tt.py", line 5, in <module>
    print(Tensor([1, 2, 3], requires_grad=True).tolist())
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/z/work/tinygrad/tinygrad/tensor.py", line 171, in __init__
    raise RuntimeError(f"Only Tensors of floating point dtype can require gradients")
RuntimeError: Only Tensors of floating point dtype can require gradients

[github-actions]

Changes

Name                  Lines    Diff    Tokens/Line    Diff
------------------  -------  ------  -------------  ------
tinygrad/tensor.py     1352      +2           20.3    -0.0


total lines changes: +2