Enable storelse for fmha_fwd_trload kernel (#3023)

Author: aska-0096

example/ck_tile/01_fmha/codegen/ops/fmha_fwd.py

@@ -608,7 +608,7 @@ def get_pipelines(dtype, hdim, hdim_v, receipt, mask_impl) -> List[FmhaFwdPipeli
                     else:
                         pipelines.append(FmhaFwdPipeline('qr_async', 'row', 't', 'f', 't', 't', logits, bias, lse, dropout, squant, mask, skip, 'f'))
                         pipelines.append(FmhaFwdPipeline('qr_async', 'row', 't', 't', 't', 't', logits, bias, lse, dropout, squant, mask, skip, 'f'))
-                        if (hdim, hdim_v) in [(64, 64), (128, 128)] and logits == "f" and bias == "no" and dropout == "f" and lse == "f" and skip == "f":
+                        if (hdim, hdim_v) in [(64, 64), (128, 128)] and logits == "f" and bias == "no" and dropout == "f" and skip == "f":
                             pipelines.append(FmhaFwdPipeline('qr_async_trload', 'row', 'f', 'f', 'f', 'f', logits, bias, lse, dropout, squant, mask, skip, 't'))
                             pipelines.append(FmhaFwdPipeline('qr_async_trload', 'row', 'f', 'f', 't', 't', logits, bias, lse, dropout, squant, mask, skip, 't'))
                     if receipt == 1 and bias != "bias":