[VC-43403] CyberArk(identity): remove backoff library and simplify error handling

[wallrj-cyberark]

The use of backoff.Retry in LoginUsernamePassword was hiding authentication errors, because the response.Success=false was not being wrapped in a backoff.Permanent error.
This caused the identity client to hang when supplied with bad credentials.

The simplest solution would have been to wrap this error:

jetstack-secure/pkg/internal/cyberark/identity/identity.go

Lines 299 to 301 in 26d6e59

	if !startAuthResponse.Success {
	return response, fmt.Errorf("got a failure response from request to start authentication: message=%q, error=%q", startAuthResponse.Message, startAuthResponse.ErrorID)
	}

I have chosen to remove the backoff.Retry altogether, because the agent already retries failed "pushes" higher up the stack.

• Eliminated the use of the backoff library for retry logic in authentication methods
• Simplified error handling by directly returning errors without wrapping them in backoff.Permanent.
• Removed unnecessary retry logic for 4xx errors in doStartAuthentication and doAdvanceAuthentication.
• Improved code readability by streamlining the LoginUsernamePassword function.

[wallrj-cyberark]

This SessionId value is important but a bit of a hack.
We want this SessionId to be passed on to the "AdvanceAuthentication" step, and causes that step to fail.
This simulates how the real identity API fails when you use an unknown username.

This change isn't related to the bad-password-hanging problem, but it allows me to write better integration tests for the LoginUsernamePassword function.

[wallrj-cyberark]

This was the root cause of the bad-password-hanging problem.
The client was retrying the start and advance authentication steps indefinitely every 10 seconds.

[wallrj-cyberark]

I removed all these backoff.Permanet wrappers because they were intended to break out of the backoff retry in LoginUsernamePassword...it was not meant to signal permanent failures to functions higher up.

[wallrj-cyberark]

These are error prefixes. The suffix contains error IDs which I didn't have time to look into. I'm not sure if the error IDs are constants or whether they are unique to each request.

[wallrj-cyberark]

I ran these tests before committing the fix and observed the tests hang until I killed them:

$ make test-unit
...
    mock.go:79: POST /Security/StartAuthentication
    identity.go:303: I0903 15:49:03.346458] made successful request to StartAuthentication source="Identity.doStartAuthentication" summary="NewPackage"
    mock.go:79: POST /Security/AdvanceAuthentication
    mock.go:79: POST /Security/StartAuthentication
    identity.go:303: I0903 15:49:13.463032] made successful request to StartAuthentication source="Identity.doStartAuthentication" summary="NewPackage"
    mock.go:79: POST /Security/AdvanceAuthentication
    mock.go:79: POST /Security/StartAuthentication
    identity.go:303: I0903 15:49:23.471064] made successful request to StartAuthentication source="Identity.doStartAuthentication" summary="NewPackage"
    mock.go:79: POST /Security/AdvanceAuthentication

...

signal: interrupt

DONE 206 tests, 2 failures in 261.042s
make: *** [make/test-unit.mk:5: test-unit] Error 130

you can see the 10s intervals between retry attempts.