Turn on lto and strip

[rsheeter]

Exploratory adjustment based on #1359.

Tested locally the impact on size and compile time is significant:

$ cargo clean && rm -rf build/ target/
$ time cargo build -p fontc --release
# branch real	2m41.823s
# main   real	0m37.618s
$ ls -l target/release/fontc | awk '{print $5" "$9}'
# branch 13048376 target/release/fontc
# main   19544344 target/release/fontc

[behdad]

What would lto = true do?

[rsheeter]

What would lto = true do?

It's an alias for fat, https://doc.rust-lang.org/rustc/codegen-options/index.html#lto. true seems less clear to me given that there are multiple significantly different ways for lto to be turned on.

[cmyr]

I agree that this would make sense for binaries that we're shipping, but less clear that it's good for all release builds? For instance crater always does a release build, and making that take two extra minutes every time has a real cost (particularly when I'm paying it to run locally on a smaller set of fonts, which might only take a couple of minutes already)

[rsheeter]

Good point. Maybe we should make the binary production for releases use a custom profile? Or ... given that the intuitive interpretation of release is that it's the one to release perhaps we should flip that and have crater use a crater-specific profile?

[cmyr]

It looks like we can have profiles inherit from other profiles, and their example is a release-lto profile: https://doc.rust-lang.org/1.85.1/cargo/reference/profiles.html#custom-profiles