Data flow: Replace hasLocationInfo with getLocation

[hvitved]

This PR changes the API of the shared data flow library from using hasLocationInfo, to instead use getLocation via the existing LocationSig signature for locations. In addition to having a nicer API, it also makes for a small performance improvement, as the 5-column hasLocationInfo predicate no longer needs to be materialized for all (Path)Nodes arising from instantiating the data flow library, we only ever need to materialize Location.hasLocationInfo.

The transformation is mostly straightforward, except for Go, where we need to account for the fact that not all existing tuples in Node.hasLocationInfo may correspond to an extracted location. This is handled by backing Location up by a newtype, which can represent both extracted and synthetic locations. In order to avoid a dependency on Node.hasLocationInfo in Locatable.getLocation, Locatable.getLocation returns not a Location, but instead a DbLocation, which can be cast to a Location, but does not extend Location.

[asgerf]

I just want to call out that this change will add more friction for the JS data flow migration, since JS data flow nodes have locations that don't correspond to any Location. I'm just wondering if this is change is unblocking something, in which case I'll just have to deal with it, but if it's mainly because it's feels "nicer" I'd appreciate it if we can try to hold off on breaking changes like this. 🙏

[hvitved]

I just want to call out that this change will add more friction for the JS data flow migration, since JS data flow nodes have locations that don't correspond to any Location. I'm just wondering if this is change is unblocking something, in which case I'll just have to deal with it, but if it's mainly because it's feels "nicer" I'd appreciate it if we can try to hold off on breaking changes like this. 🙏

As discussed offline, we agreed to go ahead with this change. The issue with missing Locations is the same in Go (see PR description).

[owen-mc]

Meant to say: the Go changes look good to me.

[aibaars]

Looks good to me. One suggestion for the TLocation type for go.

[aschackmull]

Mostly LGTM, but I wonder whether the change to source- and sink-groups is intentional? They now have no location instead of having the empty location.

[aschackmull]

I haven't looked at the Go changes, but everything else LGTM.

[tausbn]

Python 👍