Performance layer: Add new layer to measure workload performance #3

This is a new off-the-shelf layer drivers, providing a new way to consume Arm timing and hardware counter data.

This layer will allow:

• Non-invasive per frame counters (which will be noisy, but have low invasiveness).
• Serialized per frame counters (which will be less noisy, but more invasive).
• Serialized per workload counters (which will be invasive, but provide new insights we can't get through sampled counters in Arm Streamline)