Refactor EncAdaptiveLoopFilter::getPreBlkStats

[athulya-arm]

Refactor EncAdaptiveLoopFilter::getPreBlkStats

Refactor the weighted and unweighted covariance accumulation paths out
of `EncAdaptiveLoopFilter::getPreBlkStats()` into dispatchable helper
functions.

Introduce function pointers `m_getPreBlkStatsAccum` and
`m_getPreBlkStatsWeightedAccum` to `EncAdaptiveLoopFilter` class.
Move the x86 implementations into the EncoderLib x86 SIMD layer and
assign them during EncAdaptiveLoopFilter x86 initialization, while
keeping scalar helpers as the default fallback.

Also include the EncoderLib x86 sources in the VVenC CMake x86 file
lists.

Add unit tests for EncAdaptiveLoopFilter::getPreBlkStats functions.

Modify the EncAdaptiveLoopFilter constructor to include a new
'enableOpt' parameter to create a reference object and an optimized
object for testing.

Add optional `tolerance` handling to `compare_value` so unit tests can
compare results that are numerically equivalent but not bit-exact. This
allows small floating-point accuracy differences in
`getPreBlkStatsWeightedAccum` test cases. SIMD implementations can
change the order of floating-point operations compared with the scalar
reference, which may produce tiny differences while still being
mathematically correct.

[athulya-arm]

This is an initial commit that extracts EncAdaptiveLoopFilter::getPreBlkStats as a standalone function. A Neon implementation of the same will be added in a subsequent patch.

[K-os]

Hi, thank you very much for the patch. In principle, the introduced structure looks good to me, but I am not sure if there was a specific reason why the SIMD code was implemented inline here, or if maybe the added indirect function call adds too much overhead here. We'll talk about that internally and I'll get back to you.