POC PGO configuration

[cleaton]

This PR adds a POC implementation to enable rustler projects to configure scripts that will be used to optimize the rust code using PGO following: https://doc.rust-lang.org/rustc/profile-guided-optimization.html#a-complete-cargo-workflow

I've only tested it on the Struct benchmark, it seems to give small but consistent (I only tried running a few times manually) improvement especially for the bigger "decode & encode" benchmark:

Benchmarking decode ...
Benchmarking decode and encode ...
# With PGO
Name                        ips        average  deviation         median         99th %
decode                   1.06 M        0.94 μs    ±49.74%        0.92 μs        1.30 μs
decode and encode        0.80 M        1.26 μs  ±1450.35%        1.20 μs        1.46 μs

Comparison: 
decode                   1.06 M
decode and encode        0.80 M - 1.33x slower +0.31 μs


# Without PGO
Name                        ips        average  deviation         median         99th %
decode                   1.04 M        0.96 μs    ±18.04%        0.95 μs        1.28 μs
decode and encode        0.76 M        1.32 μs  ±1367.14%        1.23 μs        2.17 μs

Comparison: 
decode                   1.04 M
decode and encode        0.76 M - 1.38x slower +0.36 μs

Maybe there is a bigger rustler project we could try it on.

This was first discussed in #565

[evnu]

Should probably be somewhere in the build path to make that compatible with windows.