hotfix: bugfix to #756 (#757)

This pull request includes changes to the `flashinfer/jit/__init__.py`
file to improve the import structure and handle the `prebuilt_ops_uri`
import conditionally.

Improvements to import structure:

* Removed the unconditional import of `prebuilt_ops_uri` from
`aot_config` and added it conditionally within the try-except block to
handle cases where `_kernels` or `_kernels_sm90` are not available.
[[1]](diffhunk://#diff-39845bb8e1f81f9ca7d510e99dad0c15c7596ebe1ac909dc1b3c25b742700b5cL20)
[[2]](diffhunk://#diff-39845bb8e1f81f9ca7d510e99dad0c15c7596ebe1ac909dc1b3c25b742700b5cR48-R52)

cc @ByronHsu

Author: yzh119

flashinfer/jit/__init__.py

@@ -17,7 +17,6 @@
 # Re-export
 from .activation import gen_act_and_mul_module as gen_act_and_mul_module
 from .activation import get_act_and_mul_cu_str as get_act_and_mul_cu_str
-from .aot_config import prebuilt_ops_uri as prebuilt_ops_uri
 from .attention import gen_batch_decode_mla_module as gen_batch_decode_mla_module
 from .attention import gen_batch_decode_module as gen_batch_decode_module
 from .attention import gen_batch_prefill_module as gen_batch_prefill_module
@@ -40,13 +39,15 @@
 from .attention import get_batch_prefill_uri as get_batch_prefill_uri
 from .attention import get_single_decode_uri as get_single_decode_uri
 from .attention import get_single_prefill_uri as get_single_prefill_uri
-from .core import clear_cache_dir, load_cuda_ops    # noqa: F401
+from .core import clear_cache_dir, load_cuda_ops  # noqa: F401
 from .env import *
 from .utils import parallel_load_modules as parallel_load_modules
 
 try:
     from .. import _kernels, _kernels_sm90  # noqa: F401
+    from .aot_config import prebuilt_ops_uri as prebuilt_ops_uri
 
     has_prebuilt_ops = True
 except ImportError:
+    prebuilt_ops_uri = {}
     has_prebuilt_ops = False