Add expand option when adding environment variables

[helderco]

Fixes #5140

Note: This only applies to withEnvVariable.

Common example is to prepend to $PATH:

WithEnvVariable("VIRTUAL_ENV", "/opt/venv").
WithEnvVariable(
	"PATH",
	"${VIRTUAL_ENV}/bin:$PATH",
	dagger.ContainerWithEnvVariableOpts{
		Expand: true,
	},
).

In Python:

ctr
    .with_env_variable("VIRTUAL_ENV", "/opt/venv")
    .with_env_variable("PATH", "${VIRTUAL_ENV}/bin:$PATH", expand=True)

[helderco]

I considered keeping the utils private by moving the code from core/schema to core but decided this was simpler.

[helderco]

I've put the refactoring and the flag addition in separate commits to help in review.

[nipuna-perera]

Can this be the default behavior? IMO it's more intuitive that way. I initially assumed it worked that way. Majority of the steps of my current pipeline uses env/secret vars that need to be expanded for CLI commands so having to add the new opts for each withExec is not ideal. Could this be possibly enabled at the client level so I can set it once?

[helderco]

Can this be the default behavior? IMO it's more intuitive that way. I initially assumed it worked that way.

See @vito's reasoning in #5001 (reply in thread).

@shykes wdyt?

[helderco]

Majority of the steps of my current pipeline uses env/secret vars that need to be expanded for CLI commands so having to add the new opts for each withExec is not ideal.

@nipuna-perera This PR adds the expand option only to withEnvVariable. If you need to expand in withExec, just use a shell:

- WithExec([]string{"echo", "$PATH"})
+ WithExec([]string{"sh", "-c", "echo $PATH"})

[nipuna-perera]

Majority of the steps of my current pipeline uses env/secret vars that need to be expanded for CLI commands so having to add the new opts for each withExec is not ideal.

@nipuna-perera This PR adds the expand option only to withEnvVariable. If you need to expand in withExec, just use a shell:

- WithExec([]string{"echo", "$PATH"})
+ WithExec([]string{"sh", "-c", "echo $PATH"})

Ohh I misunderstood! Now, an expand on the withExec would also be very handy. I have quite elaborate shell commands that I have to define outside of the withExec to do string substitution. That makes the withExec less readable.