fix string and bytes convert use large memory

[ethan256]

jfr-parser consumes a lot of memory when processing large chunk files.

Part of the reason is that the read module has a lot of []byte and string interconversions.

[ethan256]

@korniltsev

before benchmark:

goos: darwin
goarch: amd64
pkg: github.com/pyroscope-io/jfr-parser/parser
cpu: Intel(R) Core(TM) i5-8257U CPU @ 1.40GHz
BenchmarkParse-8          62      18903331 ns/op     6733755 B/op     169267 allocs/op

after benchmark:

goos: darwin
goarch: amd64
pkg: github.com/pyroscope-io/jfr-parser/parser
cpu: Intel(R) Core(TM) i5-8257U CPU @ 1.40GHz
=== RUN   BenchmarkParse
BenchmarkParse
BenchmarkParse-8              63          19262599 ns/op         6552339 B/op     106477 allocs/op

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[zdyj3170101136]

you reuse the r.body, and unsafe convert it from []byte to string.

[korniltsev]

I will close this due to inactivity. This will be addressed by #22