snapshot cache: fix heartbeat panic

[danielhochman]

Encountering a panic where the snapshot has not been set yet but the cache is heartbeating and a watch comes in. Add a check on the snapshot existence for the node to avoid the panic.

Added a test to verify.

2022-08-02T15:18:27.520Z       DEBUG   v3/simple.go:391        open watch 1 for type.googleapis.com/envoy.service.runtime.v3.Runtime[rtds] from nodeID "test-cluster", version ""
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x30 pc=0xb3a163]

goroutine 7 [running]:
github.com/envoyproxy/go-control-plane/pkg/cache/v3.(*snapshotCache).sendHeartbeats(0xc0003a2180, {0xf26d10, 0xc000040030}, {0xc000496754, 0xc})
       /go/src/github.com/envoyproxy/go-control-plane/pkg/cache/v3/simple.go:193 +0x163
github.com/envoyproxy/go-control-plane/pkg/cache/v3.NewSnapshotCacheWithHeartbeating.func1()
       /go/src/github.com/envoyproxy/go-control-plane/pkg/cache/v3/simple.go:176 +0x154
created by github.com/envoyproxy/go-control-plane/pkg/cache/v3.NewSnapshotCacheWithHeartbeating
       /go/src/github.com/envoyproxy/go-control-plane/pkg/cache/v3/simple.go:167 +0x1a9

[sunjayBhatia]

LGTM

probably never hit this case in other tests since we use a longer heartbeat interval and set a snapshot before it triggers!