User password reset request responds as successful, but fails silently - TTL Issue

### problem

When a user requests a password reset at the login page, with password reset emails enabled, it returns a successful message to the user and emails them a reset token.
If the user requests another reset within that token's window (user.password.reset.ttl), it will still return a successful message to the user but willl fail internally with the message:
`DEBUG [o.a.c.u.UserPasswordResetManagerImpl] Failed to reset token and send email. Password reset token is already set for user user@emaildomain.com in domain id: xy with account primary and email user@emaildomain.com`

This doesn't appear in the event log of the user either.

### versions

4.22.1.0

### The steps to reproduce the bug

1. Have settings enabled for user password reset (user.password.reset.enabled = true) and an email host configured (user.password.reset.smtp.host)
2. Request a password reset.
3. Observe the first reset email.
4. Request another password reset.

### What to do about it?

There's a few issues here I think should be addressed:
1. Consider if password reset attemps should be visible in the user's Event's page. It might create spam in the events list, so might not be a good idea.
2. Consider either:
- Allowing a user to re-request a new password reset token and reset the TTL window using the new token (my desired outcome, and what I'd expect end-users would be familiar with).
- If not the above, then wait for the email to be successfully queued/sent and return the success from that. If the email isn't sent due to the token being within the TTL, then return a user-friendly message that they cannot request a new token until a specified datetime.

Start by reproducing the second password-reset request with user.password.reset.enabled and user.password.reset.smtp.host configured. Read UserPasswordResetManagerImpl around the logged failure and trace how the login-page response is produced. Done should be a defined, user-visible outcome when a reset token is still within its TTL, with behavior covered by the relevant password-reset tests.

Written by the indexing model from the issue text.