Fix dune subst not to initialize root

[rgrinberg]

We explicitly read --debug-backtraces instead of relying on
Common.term. Common.term ended up bringing an entire kitchen sink of
stuff we don't want. For instance, root directory discovery.

[ejgallego]

For completeness, this problem was introduced #3878 , it could make sense to have the init structure for different commands as an API in common.

[rgrinberg]

it could make sense to have the init structure for different commands as an API in common.

Indeed it would. Though I don't see any other options that are relevant to $ dune subst at the moment.

My bad for suggesting to introduce this bug in the original PR in the first place.

[ejgallego]

I don't know all the utility commands so well, thus I'm not sure what for example dune install and dune subst have in common, at least there are usually called with by opam, right? I was thinking of something such as having Common.Build_command and Common.Pkg_command etc... so we can capture the invariants of the init context of commands better.

[I am not sure I care so much about the options, but indeed more about the init process on itself]

[rgrinberg]

I don't know all the utility commands so well, thus I'm not sure what for example dune install and dune subst have in common, at least there are usually called with by opam, right?

dune install is not called by opam. Opam uses its own opam-installer binary and dune install is our own re-implementation of it.

so we can capture the invariants of the init context of commands better.

Yeah, that sounds good. I think we'll have a better idea once we add a few more commands unrelated to build.

[voodoos]

Looks good to me.

(mostly a revert of 68b32b6)