Add L1<->L2 bandwidth calculation

[skyreflectedinmirrors]

Omniperf currently does not report the achieved L2 bandwidth from the L1s, despite collecting the counters required to do so. Following the convention for L1 bandwidth calculations, this is essentially the total amount of data moved from L1<->L2, which can be calculated from the L1<->L2 requests, e.g.:

https://github.com/AMDResearch/omniperf/blob/main/src/omniperf_cli/configs/gfx90a/1600_L1_cache.yaml#L173

The L2 bandwidth calculation would be:

L2 BW = 64B * (TCP_TCC_READ_REQ_sum + TCP_TCC_WRITE_REQ_sum + TCP_TCC_ATOMIC_WITH_RET_REQ_sum + TCP_TCC_ATOMIC_WITHOUT_RET_REQ_sum) / $denom

[skyreflectedinmirrors]

It was pointed out that L1<->L2 bandwidth is a better naming scheme for this, as this excludes instruction cache and scalar cache traffic.

[coleramos425]

Is L1D Cache Accesses panel the only place we want to add this @arghdos ? To summarize, I:

1. Changed name of L1-TCR -> L1-TCC
2. Added L1-L2 BW using:

   L1_L2_BW = 64B * (TCP_TCC_READ_REQ_sum + TCP_TCC_WRITE_REQ_sum + TCP_TCC_ATOMIC_WITH_RET_REQ_sum + TCP_TCC_ATOMIC_WITHOUT_RET_REQ_sum) / $denom

Before:

After:

[skyreflectedinmirrors]

Changed name of L1-TCR -> L1-TCC

Change to match the others (L1-L2)

L1-L2 BW

I would prefer we separate the requests and the bandwidth value (right now, you do L1-L2 Read Requests, L1-L2 Bandwidth, L1-L2 Write requests...)

Additionally, make sure the the units the L1-L2 BW are Bytes per $denom, not Requests per $denom, as they are now

[coleramos425]

Merged fixes. Closing ticket.