See #931

creates the new trait layout and a first implementation

BREAKING CHANGE: new traits that replace the old

This PR changes the CI to be compatible and fitted to the new multi
crate layout of petgraph (supersedes #928).

In particular, the Testing CI is now structured as follows:
- For every applicable Rust version (stable, beta, nightly) and every
crate, runs tests with all and no-default features enabled (ci.yml)
- Runs lints on the entire workspace (ci.yml)
- Checks for no-std compatibility of the workspace (ci.yml)
- Runs every crate's tests with using cargo hack and --feature-powerset.
This runs all possible combinations of features. However, due to the
high number of features, we only check for combinations of size 1, i.e.
all features individually. This may be increased in the future, once
there are less features. This action is only run on pushes / merges to
the master branch or on PRs with the `S-run-thorough-ci-tests` label
(merge-check.yml).
- Runs cargo-semver checks and fails if there are breaking changes but
the PR is not declared as breaking. The outputs can now be found in the
workflow itself instead of posting them as comments (thus avoiding the
previous `pull_request_target` security concerns) (semver-checks.yml)
- Checks for each PR, if the title is formatted correctly. If it
contains a "!", it should also contain an explanation of the breaking
change in the body of the PR. Errors are again redirected to the
workflow output instead of posted as comments (pr-title.yml).
- There is a `miri.yml` which runs miri on the project, however this is
currently disabled, as the running time for this has increased
drastically. This needs to be investigated to be able to re-enable miri
(see #935).

Notably this PR also fixes some tests and their import of `std`
dependent objects. As a result, all tests should now pass with `std` and
`no_std`.

…ch (#952)

This tries to fix Actions not running on PRs targeting the `0.8` branch,
as seen in this PR #951.

The theory being that workflows from the `master` branch are triggered
as they are triggered on PRs to any branch but they don't actually exist
on the target branch which confuses GitHub and leads to unexpected
behaviour.

Together with gfx-rs/wgpu#9078, this should
deduplicate hashbrown for wgpu and eventually for Ruffle.

Co-authored-by: Tomáš Procházka <[email protected]>