dylib compilation and runtime loading

[enricozb]

Fixes #112.

1. Adds --dylib to the compile command. Per The Rust Reference, this will output a "*.so file on Linux, *.dylib file on macOS, and *.dll file on Window."
2. Adds --include to the run command. A space-delimited list of hvm-core dynamic library files to load before running the provided hvm-core file.
3. Fixes a bug where compiling an hvmc program that uses a single definition multiple times causes a rust error.

Additionally cleans up some clap args:

• Changes filepaths from String to PathBuf.
• Removes explicit short and long attributes if they are identical to the implicit ones.

Example:
two-and-three.hvmc:

@two = #2.0
@three = #3.0

two-plus-three.hvmc:

@main = r & @two ~ <f32.+ @three r>

Compiling and running:

$ cargo run -r -- compile two-and-three.hvmc --dylib --output two-and-three.so
$ cargo run -r -- run two-plus-three.hvmc --include two-and-three.so
#5.0
$ cargo run -r -- run two-plus-three.hvmc
thread 'main' panicked at src/host.rs:58:50:
Found reference "two", which is not in the book!
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
fish: Job 1, 'cargo run -r -- run two-plus-th…' terminated by signal SIGABRT (Abort)

Notes

Generated dylibs are versioned with the hvmc crate version that was used to compile them. Using a dylib with a version of hvmc that does not match the dylib produces a warning like the following:

warning: dylib "f32-defs.dylib" was compiled with hvmc version 0.2.25, but is being run with hvmc version 0.2.26

[HigherOrderBot]

Perf run for f28c6ac:

file              mode        main          f28c6ac98277
========================================================
sum_tree          intr-singl      10.113 s      10.021 s
                  intr-multi       4.788 s       4.740 s
--------------------------------------------------------
sum_rec           intr-singl      14.854 s      14.221 s
                  intr-multi       7.622 s       7.589 s
--------------------------------------------------------
sum_tail          intr-singl       1.230 s       1.245 s
                  intr-multi       1.188 s       1.237 s
--------------------------------------------------------
c2                intr-singl       0.000 s       0.000 s
                  intr-multi       0.001 s       0.001 s
--------------------------------------------------------
boom              intr-singl       2.231 s       2.235 s
                  intr-multi       2.282 s       2.701 s
--------------------------------------------------------
bitonic_sort_lam  intr-singl      11.539 s      11.212 s
                  intr-multi       5.692 s       5.596 s
--------------------------------------------------------
merge_sort        intr-singl       7.570 s       7.304 s
                  intr-multi       3.896 s       3.691 s
--------------------------------------------------------
radix_sort_lam    intr-singl       9.794 s       9.778 s
                  intr-multi       4.959 s       4.839 s