Interactions no longer require an execution plan

[nlordell]

Fixes regression introduced in #195 (comment)

It turns out some of our solvers were sending single interactions without an execution plan. This used to be allowed and was disallowed in the aforementioned PR, breaking backwards compatibility.

Test Plan

CI

[MartinquaXD]

I agree that we really should make that field required eventually. IMO it doesn't make sense that we should be responsible for bringing interactions in the correct order.

[fleupold]

IMO it doesn't make sense that we should be responsible for bringing interactions in the correct order.

I don't think we are at the moment neither (basically we expect the order in the array to be the correct order). But agree that making it explicit is better.

[nlordell]

@MartinquaXD: I agree that we really should make that field required eventually.
@fleupold: But agree that making it explicit is better.

Since "raw" interactions are in an array, I think it makes perfect sense for the array order to represent the order in which they are encoded. The issue arises for our internal HTTP solvers which may want to include raw interactions "inter-weaved" with AMM swaps.

This, however, is a specific problem related to the HTTP API used by our internal solvers, and I don't know if it makes sense for external solvers to have to deal with this extra "complexity" for things that it doesn't use. I think this can also be fixed by changing our HTTP API to something like:

{
  interactions: [
    { kind: "amm", /* what is in the AMM interaction today */ },
    { kind: "raw", /* what is in raw interactions today */ },
    { kind: "amm", ... },
    { kind: "raw", ... },
    { kind: "raw", ... },
    { kind: "raw", ... },
    { kind: "amm", ... },
  ],
}

This format ☝️ would also slightly more align with how settlements look like at the SC level.