design(kubernetes/remote): Sync code repository between the host and remote environment

[gaocegege]

We request users to develop on the remote container, instead of the host.

We should not assume where the user to work on

in my practice, working on remote jupyter notebook or working on local VS code, which connects to remote jupyter kernel, or working on local VS code with VS code remote development kit is the usual scenario.

on the other hand, there are a lot of algorithm engineers who do not use git as their source code management tool, they just write the code and produce some summary stuff, then the work is over, the source code does not need to manage.

file sync by envd can be a choice for the user, but can not be the only one.

how about letting the user choose how to set up their work style?

config.jupyter(password="", port=8888)
# if they want to work on a jupyter notebook
config.vscodeserver()
# or if they want to work on a VS code
config.sync(local='.', runtime='/workdir')
# or we sync their source code and datasets
# also the underhood can be a choice from volume(if the runtime is a local container)
# port-forward, nodePort, ingress, etc. if the runtime is a kubernetes

# or maybe 
envd up --sync
# but `--sync` should be the default behavior
# so, 
envd up -d --no-sync
# can be more practical
envd context ls
# then just print the context information
# if the runtime is a remote one
# tells users that VS code remote development kit can help with their work

From @TaylorHere

[gaocegege]

A per-pod sidecar can be introduced to sync files between host and Kubernetes pod. syncthing will run in the per-pod sidecar.

The sidecar design does not involve changing the image.