three (3) incorrect metrics behaviours on failed sign request

[stephengaudet]

steps to reproduce:

to observe the 3 problems, we must use a signatory service that is running in an isolated environment. take note of some metrics before and after making a failed sign request. this example uses the azure vault for the failed request.

1. curl signatory:9583/metrics and take note of the following counters:

• signing_ops_total
• vault_sign_request_error_total
• vault_sign_request_duration_milliseconds_bucket

2. make a failed signing request. with signatory started successfully, take note of the azure vault url in the config. on the signatory host, add an erroneous entry to /etc/hosts such that the azure vault url resolves to 127.0.0.1. using octez-client with signatory configured for remote signing, transfer 1 tez from the account with key in azure vault, to alice. observe octez-client report a failure

3. repeat step 1

expected:

1. signing_ops_total has not changed as no operation has been signed
2. vault_sign_request_duration_milliseconds_bucket histogram has not changed
3. vault_sign_request_error_total list contains the same error code that was logged in the signatory log

actual:

1. signing_ops_total has been incremented by the failed request
2. the failed request duration has been added to the sum in vault_sign_request_duration_milliseconds_bucket
3. signatory logs error code 500, metrics show "n/a" vault_sign_request_error_total{code="n/a",vault="Azure"}