Faster flamegraph generation

[jhawthorn]

This PR makes flamegraph generation significantly faster. The main change is to create cursor objects pointing into the raw buffer instead of slicing arrays out of the raw buffer.

Before

jhawthorn@guardian:~/src/stackprof (master ) [ruby 2.8.0]
$ time ruby -Ilib bin/stackprof --flamegraph stackprof-test-boot.dump > /dev/null
31.52s user 0.24s system 99% cpu 31.891 total

After

jhawthorn@guardian:~/src/stackprof (faster_reporting ) [ruby 2.8.0]
$ time ruby -Ilib bin/stackprof --flamegraph stackprof-test-boot.dump > /dev/null
6.69s user 0.16s system 99% cpu 6.875 total

I found this using stackprof on its own reporting 😆 ♻️

# frozen_string_literal: true

require "stackprof"

report = StackProf::Report.new(Marshal.load(IO.binread(ARGV[0])))

output = File.open("/dev/null", "wb")
StackProf.run(mode: :wall, out: 'stackprof-stackprof.dump') do
  report.print_timeline_flamegraph(output)
end

[tmm1]

This is really clever. Great use of stackprof!