Update test_blazeface.py

[thisisvk45]

Key Optimizations:

Vectorized normalization - Single operation instead of loops Cached preprocessing parameters - norm_mean, norm_std stored once Font caching - Avoid recreating fonts repeatedly
Simplified conditionals - Reduced nested if/else
Better memory management - In-place operations, reduced copies Optimized image reading - Better error handling, skip invalid files Static color map generation - Precomputed using numpy Vectorized filtering - Using boolean masks
Removed unused imports - pickle, sklearn
Cleaner code structure - Better method organization Single resize call - Directly to target size
List comprehension - For color map generation
f-strings - More efficient string formatting