test: filter pipeline e2e for grep

Unify the real-world grep tests for AMD64 and ARM64. Each test
runs a table-driven filter pipeline to verify each filter's
contribution to FP reduction, followed by full-pipeline threshold
assertions and the all/no_crt summary.

AMD64 pipeline (4 configs): none -> plt -> plt+cet -> full
ARM64 pipeline (3 configs): none -> plt -> full (CET is a no-op)

CRT functions are excluded from both ground truth and FP scoring
when evaluating against the no-CRT view.

Assertions:
- FP decreases monotonically at each pipeline step
- Full pipeline recall >= PLT-only recall
- Full pipeline recall >= 95%, FP multiplier < 0.1x

Signed-off-by: Massimiliano Giovagnoli <maxgio92@pm.me>

Author: maxgio92