zstd: Skip random data faster

[klauspost]

If we did not get a single match, assume that the data cannot be entropy compressed.

It is very likely that the data is random, and in the rare case it is not, we can live with it.

Before:

BenchmarkRandomEncodeAllFastest-12 100 10475190 ns/op 1001.01 MB/s 0 B/op 0 allocs/op
BenchmarkRandomEncodeAllDefault-12 100 11075653 ns/op 946.74 MB/s 0 B/op 0 allocs/op
BenchmarkRandomEncoderFastest-12 200 8126950 ns/op 1290.25 MB/s 139 B/op 2 allocs/op
BenchmarkRandomEncoderDefault-12 200 8059989 ns/op 1300.96 MB/s 90 B/op 2 allocs/op

After:

BenchmarkRandomEncodeAllFastest-12 300 4425967 ns/op 2369.15 MB/s 0 B/op 0 allocs/op
BenchmarkRandomEncodeAllDefault-12 300 5108791 ns/op 2052.49 MB/s 0 B/op 0 allocs/op
BenchmarkRandomEncoderFastest-12 300 5085627 ns/op 2061.84 MB/s 89 B/op 2 allocs/op
BenchmarkRandomEncoderDefault-12 300 5947031 ns/op 1763.19 MB/s 65 B/op 2 allocs/op