docs: drop fused shared experts caveat from shared-outer LoRA

Author: yueming-yuan

docs/advanced/lora.md

@@ -413,7 +413,6 @@ saving weights for sampling publishes an immutable adapter version.
 `--experts-shared-outer-loras` also works with multi-LoRA: routed experts share
 the hidden-side factor within each adapter, while tenants remain independent.
 The flag applies to the whole server and preserves the single-LoRA export format.
-SGLang currently rejects combining this layout with fused shared experts.
 
 `save_state` saves adapter parameters and optimizer state, including FP32 masters.
 It waits for preceding commands but neither steps nor saves pending gradients;