Investigate kwd performance difference between Avona and python IC

[shuchitak]

Fixes https://github.com/xmos/fwk_voice/issues/348 Partly addresses https://github.com/xmos/fwk_voice/issues/276

• Fixed an issue in the ad_state->output_energy calculation.
• Fixed counter wraparound in ic_mu_control_system()
• Fixed the acquiring of unprocessed input y frame in ic_reset_filter()
• Changed IC_INIT_MU to match python. This is used in the adaption_config = IC_ADAPTION_FORCE_ON cases and was causing kwd mismatch when I was testing with adaption set to force ON.
• Added IC+VNR stage to the python pipeline
• Added ability to override VNR predictions and IC mu in the python pipeline. This has been very useful for debugging so leaving it in.
• Replaced x86 target in pipeline test with python. Ideally, should test all 3 but would take too long to run.
• Running AEC+IC xcore and python pipeline and comparing keywords with a pass criteria of 1 in Jenkins.

[shuchitak]

With all the fixes in this PR, the AEC + IC python pipeline results between Avona and python look very similar! results_Avona_prev_arch_xcore.csv results_Avona_prev_arch_python.csv