RunMigrationsError provides no information about the specific migration that failed

[Aaron1011]

While RunMigrationsError provides information about the type of migration error that occurred, it provides no information (e.g. the file path) about the specific migration that caused the error.

Since the RunMigrationsError enum has neither a non_exhaustive attribute nor a #[doc(hidden)] variant, adding new variants is a breaking change. It was suggested on Gitter that a new error type (and corresponding functions) be created, and that the current RunMigrationsError and run_migrations be deprecated.

[sgrif]

Copying my comments from the gitter convo (and pinging the power rangers):

I can say with certainty that I intended to make that enum (and MigrationError) extensible, and it slipped through the cracks
(I wouldn't turn it into a struct, but that's semantics)
@diesel-rs/core I would like to get the rest of the core team's opinion here. I don't think there's anything we can do without a semver major version bump. I'm opposed to separating the version of diesel_migrations from the rest of the framework, and I'm also opposed to a semver-major bump on the whole shebang. That said, this enum should have been made extensible from the start. What do y'all think?

[Diggsey]

I think a couple of improvements could be made in the process: the names of some of the functions are a little odd (for example, the function to find all the migrations is called mark_migrations_in_directory) and a lot of the design seems to be quite heavily based on the implementation (for example, name() not being a method on the Migration trait, but a standalone function).

This would be a good opportunity to refactor some of this: for example creating a MigrationSet trait instead of passing around a raw PathBuf, which has methods to list Migrations.

[sgrif]

Yeah, a redesign was originally planned for 1.0 but we agreed that we could \ deprecate the current API backwards compatibly. That does not apply to types