fix(cc): compile select_map<int> when TensorFlow backend is off (#3987

)

`select_map<int>` is currently within the `BUILD_TENSORFLOW` condition
block, which is wrong. The C API uses it. This issue is found on a
machine when only building against libtorch.

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

- **New Features**
- Improved compatibility with TensorFlow by adding conditional
compilation directives.
  
- **Enhancements**
- Enhanced handling of session data with new template functions for
vectors and scalars.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Jinzhe Zeng <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

source/api_cc/src/common.cc

@@ -996,6 +996,7 @@ template void deepmd::session_get_vector<int>(std::vector<int>&,
                                               Session*,
                                               const std::string,
                                               const std::string);
+#endif
 
 template void deepmd::select_map<int>(std::vector<int>& out,
                                       const std::vector<int>& in,
@@ -1025,6 +1026,7 @@ template void deepmd::select_map_inv<int>(
     const std::vector<int>& idx_map,
     const int& stride);
 
+#ifdef BUILD_TENSORFLOW
 template float deepmd::session_get_scalar<float>(Session*,
                                                  const std::string,
                                                  const std::string);