The any_env.py and any_env2.py recipes were installing pandas==1.5.3 without
constraints, which caused pip to install numpy 2.4.1 (latest). This resulted
in binary incompatibility:

  ValueError: numpy.dtype size changed, may indicate binary incompatibility.
  Expected 96 from C header, got 88 from PyObject

Root Cause:
- Recipes have embedded create_data_popen() function that generates virtualenv
  creation scripts
- The pip install command for packages in modules_needed_by_name was:
    {env}/bin/pip install {pkg} --ignore-installed
- This doesn't constrain dependencies, so pandas pulls latest numpy (2.4.1)
- pandas 1.5.3 was compiled against numpy 1.x, incompatible with 2.x

Fix:
- Add req_constraints_deps.txt to pip install command:
    {env}/bin/pip install --use-deprecated=legacy-resolver -c {constraints} {pkg} --ignore-installed
- This ensures numpy 1.24.4 is used with pandas 1.5.3, maintaining binary compatibility

This matches the fix already applied to h2oaicore/utils.py in the main DAI repo.

Fixes: test_contrib.test_pools.test_repo_download_specific failure

This reverts commit 32ab3b4.