[gfx1250] feat(mega_moe): add operand init modes and b2b timing knobs

[arakowsk-amd]

Summary

• Add a shared _init_data helper supporting zero / const / rand (uniform) / norm (gaussian) operand initialization, with rand/norm honoring the generator for reproducibility.
• Wire _init_data into make_shared_weights and the x0 activation tensor; expose --init and --init-const CLI flags.
• Add --warmup and --iters knobs so DeviceMoEPipeline.bench() can run averaged back-to-back (b2b) graph replays with a single sync at the end, instead of a single fixed replay.

Test plan

• python -m py_compile op_tests/multigpu_tests/test_mega_moe_gfx1250.py (passes locally)
• Run the mega_moe multi-GPU benchmark on gfx1250 with default args (regression: unchanged --init norm --iters 1 path)
• Exercise new flags: --init const --init-const 1.0, --init zero, --iters 4 --warmup 10

🤖 Generated with Claude Code

[github-actions]

🏷️ CI Guide

Runs automatically on every PR:

• ✅ Pre-checks (submodule verification, code formatting)
• ✅ Aiter op tests (gfx942 + gfx950)
• ✅ Triton tests on MI35X (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
ci:gfx1250-ffm-triton	Run the five-shard gfx1250 FFM Triton test suite
ci:triton-300x	Run an additional Triton test job on MI300X in PRs; main branch always runs both MI35X and MI300X
multigpu	Aiter multi-GPU tests on the 8-GPU runner
ci:sglang	SGLang integration tests: DeepSeek-R1-MXFP4 accuracy, Qwen 3.5 accuracy
ci:atom	ATOM benchmark: DeepSeek-R1-0528, GPT-OSS-120B
ci:atom_full	ATOM accuracy suite for PR and main models from ATOM models_accuracy.json
ci:vllm	vLLM benchmark: GPT-OSS-120B, DeepSeek-R1-0528, Kimi-K2.5
ci:all	All standard extended tests (excludes ci:atom_full)

Only add ci:atom_full for FlyDSL or Triton upgrades.
Add labels via the sidebar or gh pr edit 5163 --add-label <label>

PR title tags & labels:
Component tags ([Triton/Gluon], [HIP], [CK], [ASM], ...) are added to the PR title and as PR labels automatically from the changed files and re-synced on every push — change-type tags like [fix]/[Perf], op tags like [MLA], and human labels (ci:*) are left untouched. Add the no-auto-title label to opt this PR out.