This changes adds a --cuda-compat-mode flag to the configure
CLI. This allows more flexibility than the existing --no-cntlibs flag.

Possible values of the flag are:
* mount (default) - CUDA compat libraries are mounted from /usr/local/cuda/compat to
  the standard library path in the container.
* ldconfig - The folder containing the CUDA compat libraries is added as a command
  line argument to the ldconfig command executed in the container.
* disabled - This is equivalent ot specifying the --no-cntlibs flag and skips
  the detection and injection of compat libraries from the container to the
  container entirely.

Signed-off-by: Evan Lezar <elezar@nvidia.com>

Since the compat libraries are discovered as part of container
creation, the --no-cntlibs option that disables their discovery
is not relevant to the nvidia-container-cli list command.

This change removes this flag.

Signed-off-by: Evan Lezar <elezar@nvidia.com>

Add cuda-compat-mode flag to configure command