Right now @devcontainers/cli is executed using the agentexecer package. However, some environment manipulation is done in (*agentssh.Server).CreateCommand. We should replicate this environment when we run devcontainer up/read-configuration/exec or docker.

This is necessary to ensure that we look for commands in the correct paths, or any other special environment variables present that are required for the aforementioned commands to work.

While we're at it, we don't really need the Docker env infoer with the current approach and could remove it in favor of our current use of devcontainer exec. This should help reduce dependencies to CreateCommand.