Add RTX 3060 Laptop support

[AndrewVu23]

Summary

Wires the RTX 3060 Laptop into Accel-Sim end to end: config registry entry, correlation mapping, the HIT + WRITE_ALLOCATED comparable write-hit metric, DP 64,64 trace config, and an NVBit 1.7.6→1.8 bump required for driver ≥59x.

Motivation / root cause

NVBit 1.7.6 corrupts traces on driver 596 (garbage register operands -> post-processing crash); NVBit 1.8 fixes it.
The correlator had no RTX3060 mapping, and its L2 write-hit stat used strict HIT (see related 2)

Changes

• define-standard-cfgs.yml: RTX3060M base-config entry.
• correl_mappings.py: RTX3060M card map; L2 Write Hits/Rate = HIT + WRITE_ALLOCATED.
• example_stats.yml: collect WRITE_ALLOCATED.
• install_nvbit.sh: NVBIT_VERSION=1.8.
• trace.config: DP 64,64 (latency ≥ init; consumer-Ampere 1/64 DP) (see related 1)

Evidence (before / after)

Before: corrupt trace lines / post-process crash on driver 596

After: 0 corrupt lines across 16 ubenchs. Correlation: full ubench + 10-app rodinia (GPC 0.997, L2 write-hit 0.999 comparable)

Test plan

• run_hw_trace.py produces clean traces (NVBit 1.8)
• get_stats.py -R -K -k + plot-correlation.py produce the dashboard
• correlation runs against a binary built from (see related 2)

Compatibility / backward compatibility

• NVBit 1.8 doesn't interfere with any existing framework. Run perfectly fine.
• New config + additive suites. The correl_mappings.py write-hit change is coupled to (see related 2)

Related

1. Guard pipelined_simd_unit against initiation_interval > latency gpgpu-sim_distribution#139
2. Add additive L2 stat for NVIDIA-comparable write hits (GAME CHANGER!!!) gpgpu-sim_distribution#141