Upgrade google.golang.org/grpc, fixing gogoproto incompatibility

[stevendanna]

In a different line of work, attempting to upgrade a client library for
google cloud APIs brought in an upgrade to google.golang.org/grpc and revealed
that a recent changes to google.golang.org/grpc have modified the default
proto codec such that it is incompatible with gogoproto:

grpc/grpc-go#3958

While proto.Marshal does do a check for the Marshal/Unmarshal it
appears to do this through reflection methods that will panic on
fields using gogoproto.nullable.

To avoid this, we install a proto codec that does an interface check
before dispatching to proto.Marshal/proto.Unmarshal.

Note that I haven't replicated the protobuffer caching that previously
existed in google.golang.org/grpc. The referenced change claims
removing the caching may be a small performance win; however, they
also removed the typecheck that we need to do, so it is unclear.

The upgrade of google.golang.org/grpc is larger than I might like, but I
was unable to convince Go to upgrade fewer dependencies as the grpc update
forces an update of google.golang.org/api, which then forces an upgrade
of a number of other dependencies. If we don't want to take the dependency
upgrade yet, we can merge the first commit to unblock future upgrades.

Fixes #60531

[cockroach-teamcity]

This change is 

[tbg]

Thank you @stevendanna and sorry for letting this sit! Due to the stability period, I would postpone the merge to after the release branch has been cut. I have held off on scrutinizing the dep bump so far, so we'd have to do that before the merge.

Also cc @knz for server/networking.

Reviewed 4 of 4 files at r1, 8 of 8 files at r2.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @irfansharif)

[knz]

wowzer that's a lot of dep changes.

The changes inside crdb LGTM, but as usual we will want to scrutinize the upstream changes as well. Agreed with Tobias that may not be suitable for the 21.1 release but we should get this started soon afterwards.

Reviewed 3 of 4 files at r1, 7 of 8 files at r2.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @irfansharif)

[stevendanna]

@knz @tbg Completely understandable and likely very prudent.

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @irfansharif)

---

pkg/rpc/codec.go, line 29 at r2 (raw file):

		return pm.Marshal()
	}
	return protoutil.Marshal(v.(protoutil.Message))

I updated this line to use protoutil in accordance with the linter. But, I think we should probably exclude this file from the linter and move this back to calling one of the proto libraries directly. My concern is that the conversion to protoutil.Message requires that the MarshalTo method be implemented which I don't think is going to be the case for various third party libraries.

[stevendanna]

I've pulled out the first commit and added some tests for it here: #61333

I still think both the new codec and this upgrade can wait until we cut a new branch though.