Fixes #62 by implementing #62 (comment)

Schemars currently differentiates between types using the JsonSchema::schema_name() function, which typically returns the type name (excluding the module path). This PR updates the schema generator to differentiate between types using the new function JsonSchema::schema_id(), which is similar, but includes the type's module path. This leaves schema_name() returning a clean, human-readable "friendly" name mostly free of rust details (e.g. being unaffected by module renames).

schema_id() has a default implementation (which delegates to schema_name()), so this should be a non-breaking change.

Because a vast majority of types have a statically-determined id (and name), schema_id() returns a Cow<'static, str> instead of a String - in future, I'd like to change schema_name() to return a Cow for the same reason, but that will be a breaking change so is not being done in this PR. The main reason an implementation would return a Cow::Owned(String) would be a generic type, e.g. Vec<T>, where the return value of schema_id() (and of schema_name()) depends on that of T.

In many cases, the return value of T::schema_id() will be identical to that of std::any::type_name::<T>(), so ideally we could just use that - unfortunately type_name() is unsuitable here for two reasons:

1. the output of type_name() may change in a future rust version e.g. by excluding the module path, or including lifetime specifiers
2. some distinct rust types have equivalent schemas, so should have identical return values of schema_id(), e.g. String/str/&str, but these will all have different return values of type_name()

Still to do:

• Update docs
  • doc comment for schema_id() ✅
  • web docs - will be done in a separate change
• implement schema_id() appropriately for optional dependencies ✅
• add tests ✅