Fix Transaction ID counter for Postgres backups

[jzelinskie]

Problem Statement

When performing a backup of a Postgres database, Postgres does not backup/restore the transaction ID counter.
Because SpiceDB depends on this value being set, we need some means of setting the counter or incrementing it to the place where the previous database counter was set.

Solution Brainstorm

After some investigation @jakedt was able to produce a benchmark for batching the creation of transactions to increment the Postgres transaction ID counter. With batch size of 10k its ~2.6 seconds per million transactions. This could probably use a bit more investigation to find an ideal batch size.

func BenchmarkTxidPipeline(b *testing.B) {
	ctx := context.Background()

	conn, err := pgx.Connect(ctx, localhostDB)
	if err != nil {
		b.Fatalf("err: %s", err)
	}
	defer conn.Close(ctx)

	b.ResetTimer()

	// b.Logf("b.N: %d", b.N)
	for i := 0; i < b.N; {
		var batch pgx.Batch

		for j := 0; j < batchSize && i+j < b.N; j++ {
			// b.Logf("queuing query: %d", i+j)
			batch.Queue("begin;")
			batch.Queue("select pg_current_xact_id();")
			batch.Queue("rollback;")
			executed++
		}

		// b.Log("sending batch")

		i += batch.Len()

		br := conn.SendBatch(ctx, &batch)

		if err := br.Close(); err != nil {
			b.Fatalf("err: %s", err)
		}

		// b.Logf("value of i: %d", i)
	}

	b.Logf("total executed: %d", executed)
}

BenchmarkTxid-10            	     170	   6499092 ns/op
BenchmarkTxidPipeline-10    	  400071	      2705 ns/op

After this explored, we need to introduce this into SpiceDB.
A "repair" command could be added as maybe a part of a larger command for sanity-checking configurations (e.g. spicedb status --repair)