feat: /flex-mode command

https://github.com/openai/codex/pull/372 adds support for --flex-mode to allow the user to specify at launch whether they want to use the flex or default service_tier, as requested in https://github.com/openai/codex/issues/339.

If users are in the middle of an interactive codex session and flex-mode starts getting slow, they may want to deactivate it. Conversely, if they and ready to step away from the session for awhile and want codex to proceed at an economical pace, they may want to enable it.

The easiest-to-implement method for activating and deactivating flex-mode within a session is to add a /flex-mode command.