Use floating-point literals in TrackUtils.h

[f3sch]

This shows up during profiling ITS GPU kernels.
The division on line 176 is forced to be calculated in double. This costed 23% of stalls in some kernels.
I am not sure if the precision is required here?
At least in local tests it did not affect the ITS tracking output (<0.1%).

[davidrohr]

That was definitely a bug, and is a common case which can force GPUs to do double arithmetics instead of single.
Though, I am wondering do we need to cast everything to value_t? In my opinion, it makes the code a bit more complicated. Changing 1. to 1.f should be enough?
All the constants are anyhow only float precision, so no reason to cast to double in case value_t is double.

Obviously this is a matter of taste, but in my code I prefer to use use float constants.
Of course, in case we want to support value_t = half or so, the version of this PR is better.

[f3sch]

Hi David, that's how I found it. I profiled ITS kernels which were forced to do double here. I first did I like you suggested 1.f and Sandro suggested to use value_T. I honestly do not have a preference here.

[davidrohr]

If that was intentional, one should put a comment, looks like an oversight.

OK, I don't care too much for the style if it is outside my code :).