NVFuser kernels: floating point scalars interpreted as `Double` and not `Float`

@kevinstephano is already working on this.

Since Nov 12, 2024.

We have primitives that are inconsistent when it comes to interpreting input floating point scalar values.
For example, take a look at `where` (relevant issue https://github.com/Lightning-AI/lightning-thunder/issues/833), where we can see a result-meta mismatch.
Also, take a look at the fused section from https://github.com/Lightning-AI/lightning-thunder/issues/833:
```python
def nvfuser_fusion_id0(fd : FusionDefinition) -> None :
T0 = fd.define_tensor(shape=[-1, -1], contiguity=[True, True], dtype=DataType.Float, is_cpu=False, stride_order=[1, 0])
S1 = fd.define_scalar(0.00000, dtype=DataType.Double)
T2 = fd.ops.gt(T0, S1)
S3 = fd.define_scalar(-0.00000, dtype=DataType.Double)
S4 = fd.define_scalar(0.00000, dtype=DataType.Double)
T5 = fd.ops.where(T2, S3, S4)
fd.add_output(T5)
```
So, we should normalize scalar interpretation across executors.

cc @tfogal