refactor(edit): Switch to toml_parse

[epage]

We don't expect much difference with this change:

• This helps ensure toml_parse is as valid as possible
• This paves the way for toml to depend on toml_parse, rather than toml_edit
• Error messages are changed
• This unblocks us from offering multiple parse errors at once

The quality of error messages is a mixed bag. Some improved while others got worse. Overall, the worse ones should be the less common ones.

Despite passing the TOML conformance tests, this is a risky change

• It has less reuse between different parts of the parser, so if any tests assume a feature is covered by another test, it might not be the case
• Each layer makes assumptions about validation at the other layers making auditing difficult

Performance

To reduce extra jitter from allocators being hammered, I ran each benchmark on its own. Each benchmark was run about 10 times and I pulled out the fastest result.

$ cargo bench --bench 0-cargo -- toml::document::2-features

before: 738us
after: 686us (-7%)

$ cargo bench --bench 0-cargo -- toml::manifest::2-features

before: 646us
after: 589us (-9%)

$ cargo bench --bench 0-cargo -- toml_edit::document::2-features

before: 527us
after: 477us (-9%)

$ cargo bench --bench 0-cargo -- toml_edit::manifest::2-features

before: 650us
after:593us (-9%)

$ cargo bench --bench 1-map -- toml::document::100

before: 75us
after: 57us (-24%)

$ cargo bench --bench 1-map -- toml_edit::document::100

before: 63us
after:46us (-27%)

$ cargo bench --bench 2-array -- toml::document::100

before: 65us
after:46us (-29%)

$ cargo bench --bench 2-array -- toml_edit::document::100

before: 61us
after: 41us (-33%)

EDIT:: I forgot to run these numbers with unsafe or perf features enabled on toml_edit

[github-advanced-security]

clippy found more than 20 potential problems in the proposed changes. Check the Files changed tab for more details.