refactor: test error semantics

[ARR4N]

Why this should be merged

Testing errors by matching their strings results in change-detector tests that can result in false positives due to nothing more than changing the message intended for humans. Changing an implementation and a test at the same time adds uncertainty to the correctness of the change, but change-detector tests force this to be done.

See also: Test error semantics

How this works

The errs package is introduced to allow tagging of errors with distinct identifiers that are intended to be immutable (i.e. no change-detector tests) even when messages change. The IDs can only be extracted via the errs package, which is in internal to stop leakage of the errors via the API such that users can depend on them as per Hyrum's Law:

With a sufficient number of users of an API,
it does not matter what you promise in the contract:
all observable behaviors of your system
will be depended on by somebody.

How this was tested

Unit test for ID extraction + usage in params tests.