Test solver behavior when using options

[jrray]

It is expected that the solver will make an effort to resolve packages that fit the options, even though it is not required to. Create a test where multiple builds exist with different values for a var and add an option to the solver for a specific value. A build exists with that value so the solver should pick it.

[jrray]

I labeled this with bug because the step solver doesn't pass the test. First things first, do we believe this test is valid and should pass?

[jrray]

spk/crates/spk-cli/common/src/flags.rs

Line 291 in 2946a9a

solver.update_options(option_map);

This is the only place in the codebase (outside of test code) that calls Solver::update_options. The test suite still passes if this line is removed (on the main branch).

The new test in this PR demonstrates that setting options, at least in this trivial case, doesn't seem to do anything. Using --opt on the command line turns those things into var requests. We don't seem to be negatively impacted by the fact that this new test doesn't pass. Is it worth the effort to try to make it pass? What if instead we remove the concept of options? We keep var requests and the --opt flag and all the current behavior stays the same.

[rydrman]

From the meeting today:

• the only concern here is whether it would have an impact on build ordering (seems like now) but related to that is with source builds
• likely we can remove this part of the solver interface altogether, but we do still want the solution to contain the options from the resolve