[fix][SWDEV-413305][nvcc] Recovering the fallback building of HIP programs via nvcc

[Synopsis]
+ There is no support for an explicit `-x cu` in hipcc anymore
+ Thus, the building via `nvcc` relies only on `HIP_PLATFORM`, which should be set to 'nvidia' in order to invoke `nvcc` compilation with implicit providing `-x cu`
+ But such an implicit setting of `-x cu` never happens
+ As a result, the compilation via `nvcc` starts actually, but ends up with the following error:
`nvcc fatal   : Don't know what to do with 'MatrixTranspose.hip'`

[Solution]
+ Not taking into account the excessive variable `$hasCXX` in the condition for an implicit setting of `-x cu`
+ [Reason] `$hasCXX` is never set for the `nvcc` path

[Testing]
1. [Windows + MS cl as host compiler]
Steps:
- set HIP_PLATFORM=nvidia
- set HIP_PATH=<path to ROCM HIP SDK>
- set CUDA_PATH=<path to NVIDIA CUDA Toolkit>
- set set path=c:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.36.32532\bin\Hostx64\x64\;%path%
- perl hipcc.pl MatrixTranspose.hip -o MatrixTranspose -v --gpu-architecture=sm_XX
- run MatrixTranspose.exe

[NOTES]
- The path to the MS cl compiler is needed cause nvcc uses it for host compilation
- Setting the `--gpu-architecture` depends on NVIDIA GPU to run on
- The execution results are correct according to the margin of error = 0.000001

2. [Linux] To be accomplished by PSDB

Author: emankov

amd/hipcc/bin/hipcc.pl

@@ -532,7 +532,7 @@ sub get_normalized_path {
     $ENV{HCC_EXTRA_LIBRARIES}="\n";
 }
 
-if ($hasCXX and $HIP_PLATFORM eq 'nvidia') {
+if ($HIP_PLATFORM eq 'nvidia') {
     $HIPCXXFLAGS .= " -x cu";
 }