Underlying error causing timeout not properly surfaced in normal/unary call retries

[leahecole]

See internal bug 351328276 for more context. Based on this bug, I believe that we throw an error when the timeout is exceeded as part of the "normalCalls" retry cycle (https://github.com/googleapis/gax-nodejs/blob/d74fa409897468af5dba78ef60bb5d4c9defacb6/gax/src/normalCalls/retries.ts#L810) but it seems like the internal error that it was retrying on before timing out may be getting swallowed up by that retry cycle.

Steps to fix:

• On timeout, add logic to make sure error is surfaced to the caller
• Update unit tests (likely this one and similar ones around it, or add new ones)
• Add a test or two or three to the test application. Might be a modification of echo error and/or could be a chance to write some tests against the regular (non-streaming) methods of the sequence client - this would look similar to some of our streaming tests, but would utilize CreateSequence and AttemptSequence to send a specific sequence of errors rather than CreateStreamingSequence and AttemptStreamingSequence