unify path types in combinator, segutil, sciond

[matzf]

• Combine mostly equivalent implementations of snet.Path
  segutil.path and sciond.Path, in a shared lib in go/lib/snet/path.

• Adapt combinator to return a Path that is already almost identical
  with this path -- the only missing bit is the UnderlayNextHop (which is
  outside of the scope of the combinator, not having access to the local
  topology).
  In particular, the combinator now directly builds a raw forwarding path
  instead of returning a high level description of the same -- using the
  structs of the slayers for this actually allows to get rid of most of the
  intermediate helper types.
  This makes the places that call the combinator fairly trivial.

• Side quest: strip down the public interface of the combinator to the
  essentials (combinator.Combine and combinator.Path).

• Drop sciond.FwdPathMeta etc, use snet.Path directly
  in the sciond grpc.Daemon server. This is nice because the "input" on the
  server side and the output on the client side are now the same type.

• Drop various unused sciond helper types

• Remove sciond staticinfo "condenser" and seperate types. Instead of
  this, the staticinfo metadata will be added to snet.PathMetadata.
  This data will be prepared and returned in the identical format directly
  by the path combinator -- this is the motivation for this refactoring!

• Remove unused sciond/fetcher/filter (instead of fixing the tests)

Planned next steps (future PRs):

• add the "staticinfo" information to the snet.PathMetadata, adapt the existing (currently unused) code in the combinator to fill this information, adapt the sciond grpc proto to carry this extended metadata.
• change snet.PathMetadata from an interface to a struct. Similar to the PathInterface, there simply seems no reason for this to be an interface.
  Depending on how exactly the extended "staticinfo" data is represented, this might be useful to allow defining higher-level accessor functions to aggregate the information but still giving applications the full information (e.g. storing latency for each hop, but allowing a simple TotalLatency() getter) -- note that having to deal with missing information may require the applications to have more insight than only the pre-condensed data.
  Perhaps move the Interfaces list into this Metadata struct as well, as it seems to belong in here.
• perhaps, attempt to turn snet.Path into a struct as well -- it seems that the behaviour of all the different implementations (empty path, partial path, ...) can easily be combined.
• remove the now pointless segutil.Router

Closes #3872

---

This change is 

[lukedirtwalker]

Reviewed 27 of 27 files at r1.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on @matzf)

a discussion (no related file):
nice work 💯

---

---

go/cs/segutil/router.go, line 26 at r1 (raw file):

// Router returns paths backed by the local path database.
// XXX(matzf): this doesn't do meaningful work anymore, drop it.

💯

---

go/lib/snet/path/path.go, line 39 at r1 (raw file):

	SPath   *spath.Path
	NextHop *net.UDPAddr
	IFaces  []snet.PathInterface

not super happy with this naming, but once we move it to PathMetadata we might be able to have a clean name, no method needed.

---

go/lib/snet/path/path.go, line 81 at r1 (raw file):

		return p.Dst
	}
	return p.IFaces[len(p.IFaces)-1].IA

don't we always set Dst correctly?

---

go/pkg/sciond/fetcher/fetcher.go, line 114 at r1 (raw file):

Bad

I know this was as before, but it should be lowercase bad as error messages should start with lower case. (https://github.com/golang/go/wiki/CodeReviewComments#error-strings )

[matzf]

Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @lukedirtwalker)

---

go/lib/snet/path/path.go, line 39 at r1 (raw file):

Previously, lukedirtwalker (Lukas Vogel) wrote…

not super happy with this naming, but once we move it to PathMetadata we might be able to have a clean name, no method needed.

Ack.

---

go/lib/snet/path/path.go, line 81 at r1 (raw file):

Previously, lukedirtwalker (Lukas Vogel) wrote…

don't we always set Dst correctly?

Indeed! Done.

---

go/pkg/sciond/fetcher/fetcher.go, line 114 at r1 (raw file):

Previously, lukedirtwalker (Lukas Vogel) wrote…

Bad

I know this was as before, but it should be lowercase bad as error messages should start with lower case. (https://github.com/golang/go/wiki/CodeReviewComments#error-strings )

Done.

[lukedirtwalker]

Reviewed 4 of 4 files at r2.
Reviewable status: complete! all files reviewed, all discussions resolved

[lukedirtwalker]

/rebase