B5 coverage update

[coreyqh]

WIP

[Lamarr3314]

MinNorm +- 3ulp corrections:
The current coverpoint on B5 checks the bit just before the mantissa. Because there is padding included before the intermediate results, this means that the padded 0 will result in incorrect coverage for values where the mantissa must be all 1s. While this technically works for +3ulp because all +3ulp values have a string of 0s and that padded 0 does not affect it, we should have consistency and change the range for both +-3ulp.
F32 + 3ulp
F32 - 3ulp
F64 + 3ulp
F64 - 3ulp
F128 + 3ulp
F128 - 3ulp
F16 + 3ulp
F16 - 3ulp
BF16 + 3ulp
BF16 - 3ulp
MinSubNorm +-3ulp corrections:
Expand Range of the mantissa to also have the lsb included. For example, the if statement for F32 + 3ulp values should read: iff (CFI.intermX == 0 && CFI.intermM[INTERM_M_BITS -: F32_M_BITS + 1] == 1) {
F32 + 3ulp
F32 - 3ulp
F64 + 3ulp
F64 - 3ulp
F128 + 3ulp
F128 - 3ulp
F16 + 3ulp
F16 - 3ulp
BF16 + 3ulp
BF16 - 3ulp
Additionally, there is a coverpoint called minSubNorm_pm_3_ulp which is unused. Correcting the above +-3ulp coverpoints above should be sufficient. these unused pm coverpoints can be removed.
F32_pm_3ulp
F64_pm_3ulp
F12_pm_3ulp
F16_pm_3ulp
BF16_pm_3ulp
Also, in your coverage, you reference a coverpoint called "precision"_btw_minSubnorm_zero. This coverpoint does not exist anywhere in your code. Making this function will allow the btw_minSunNorm_zero check to work.
F32_btw
F64_btw
F128_btw
F16_btw
BF16_btw

[coreyqh]

@Lamarr3314 Arithmetic operations should be good now. I will get to converts later tonight.

[coreyqh]

Converts are annoying. Mostly done here, just need to write a few helper coverpoints

[coreyqh]

@Lamarr3314 The covergroup should be good to go. Let me know if you see anything dubious, otherwise I will merge in the morning.