Add flop calculation for custom Qwen 3 model

[gagika]

Description

Adds a QWEN3_CUSTOM_MOE branch to the FLOPs calculator that accounts for the compressed moe_expert_input_dim / attention_output_dim (768) used by the FFN/gate and attention projection, plus the per-layer up-projection back to emb_dim.

Tests

• test_qwen3_custom_30b_a3b_flops validates against the 6·N·D
  heuristic (~2.86B active params)

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.