Replace cli flag --prebuilt-platform with --build-host

[lukewilliamboswell]

This PR decouples roc from the host toolchains, by moving host rebuilding behind a compiler flag so it is no longer the default behaviour. It is still available where required using --build-host however will give a warning;

WARNING: platforms are responsible for building hosts, this flag will be removed when internal test platforms have a build script

This PR effectively removes platform rebuilding from the cli for end users -- while keeping some rebuilding functionality just for internal cli tests.

Application authors who use a release of a platform are not affected by this change. These users will continue to use a prebuilt platform.

After this change, platform authors will need to include a build script in their platform to build host binaries.

This change will make running platform example less convenient if they are not using a URL release, as the platform binaries will need to be built first in a separate step. For example this workflow;

1. Build the platform binary roc build.roc using host toolchain e.g. cargo or zig
2. Run the example application roc examples/hello-world.roc

[lukewilliamboswell]

For this CI failure in nix-ubuntu... there is something going on between Nix and basic-cli. I'm going to ignore these tests (for linux) for this PR because we will be removing these in a follow-up as they are the only remaining basic-cli dependencies and prevent roc from moving ahead of the platform.

stack backtrace:
   0: rust_begin_unwind
             at /rustc/bf3c6c5bed498f41ad815641319a1ad9bcecb8e8/library/std/src/panicking.rs:647:5
   1: core::panicking::panic_fmt
             at /rustc/bf3c6c5bed498f41ad815641319a1ad9bcecb8e8/library/core/src/panicking.rs:72:14
   2: assert_no_unexpected_error
             at /home/runner/work/roc/roc/crates/cli_test_utils/src/command.rs:137:5
   3: assert_success_with_no_unexpected_errors
             at /home/runner/work/roc/roc/crates/cli_test_utils/src/command.rs:83:9
   4: cli_test_utils::command::CmdOut::assert_clean_success
             at /home/runner/work/roc/roc/crates/cli_test_utils/src/command.rs:68:9
   5: check_build_and_run
             at /home/runner/work/roc/roc/crates/cli_test_utils/src/exec_cli.rs:84:9
   6: cli_tests::cli_tests::test_platform_basic_cli::module_params_pass_task
             at ./tests/cli_tests.rs:436:13
   7: cli_tests::cli_tests::test_platform_basic_cli::module_params_pass_task::{{closure}}
             at ./tests/cli_tests.rs:425:37
   8: core::ops::function::FnOnce::call_once
             at /rustc/bf3c6c5bed498f41ad815641319a1ad9bcecb8e8/library/core/src/ops/function.rs:250:5
   9: core::ops::function::FnOnce::call_once
             at /rustc/bf3c6c5bed498f41ad815641319a1ad9bcecb8e8/library/core/src/ops/function.rs:250:5
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.


failures:
    cli_tests::test_platform_basic_cli::combine_tasks_with_record_builder
    cli_tests::test_platform_basic_cli::module_params_different_types
    cli_tests::test_platform_basic_cli::module_params_pass_task

[smores56]

Some minor stuff, but the shape is good.

This PR is a behemoth and has been cooking for many months. I'd vote that we continue to address feedback in a different PR, as every time I come back to this PR after a merge request, thousands of lines have to be re-read or re-skimmed.

That said, this was a Herculean task and everyone involved in this PR should be proud of their work!

[Anton-4]

I'd vote that we continue to address feedback in a different PR

I already started with some cleanup before I read this, CI should be good though 🤞

[Anton-4]

Much love and gratitude @lukewilliamboswell @smores56 @r-bar ❤️