discrepancy of results between different compilers

Issue likely due to the use of aggressive floating point optimization (fast-math) set at the top CMakeList, impacting all float operations in VTM encoder. Possible fixes:

• try to restrict the flag to SADL files: might require to isolate SADL files in cpp unit
• remove completely the flag: might impact the SADL float model performance
• remove code with float operations in the encoder