Added a Next method to the Lexer (which avoids allocations) as an alternative to Scan (which provides more debugging info)

[awalterschulze]

Added a benchmark to show how there are now no allocations as part of Next, see example/mail/parser_test.go

mail % go test -v --test.bench=.
=== RUN   TestScan1
--- PASS: TestScan1 (0.00s)
=== RUN   TestNext1
--- PASS: TestNext1 (0.00s)
=== RUN   TestScan2
--- PASS: TestScan2 (0.00s)
=== RUN   TestNext2
--- PASS: TestNext2 (0.00s)
goos: darwin
goarch: arm64
pkg: github.com/goccmack/gocc/example/mail
cpu: Apple M4 Pro
BenchmarkScan
BenchmarkScan-14         3863421               294.4 ns/op           240 B/op          3 allocs/op
BenchmarkNext
BenchmarkNext-14         5198650               234.3 ns/op             0 B/op          0 allocs/op
PASS

The idea is that this can be used for high performance validation

[awalterschulze]

This is not necessary we can just copy these methods beside the Lexer for faster lexing, we do not need to generate them.