fix: handle ActiveRecord::RecordInvalid in ssh_creds when workspace is reset

[4ravind-b]

Fixes #21480

When workspace -D is run while a session is active, the session's
database record gets deleted but the session itself stays alive. When
post/multi/gather/ssh_creds tries to save credentials, it fails with
ActiveRecord::RecordInvalid: Session can't be blank and crashes the module.

This fix adds a rescue for ActiveRecord::RecordInvalid so the module
prints a warning and continues instead of crashing.

Verification

• Start msfconsole
• Get a session on target
• Run post/multi/gather/ssh_creds — works fine
• Run workspace -D
• Run post/multi/gather/ssh_creds again
• Verify module prints a warning instead of crashing
• Verify files are still downloaded successfully

[4ravind-b]

Verified on Metasploitable2.

Before this fix: Running the module after deleting the workspace with workspace -D caused the module to crash and print a full stack trace.

After this fix:

msf post(multi/gather/ssh_creds) > workspace -D
[*] Deleted workspace: default
[*] Recreated the default workspace

msf post(multi/gather/ssh_creds) > run
[*] Finding .ssh directories
[*] Looting 3 .ssh directories
[*] Looting /home/msfadmin/.ssh directory
[+] Downloaded /home/msfadmin/.ssh/authorized_keys -> ...
[+] Downloaded /home/msfadmin/.ssh/id_rsa -> ...
[!] Could not save credential to database: Validation failed: Session can't be blank
[+] Downloaded /home/msfadmin/.ssh/id_rsa.pub -> ...
[*] Looting /home/user/.ssh directory
[+] Downloaded /home/user/.ssh/id_dsa -> ...
[!] Could not save credential to database: Validation failed: Session can't be blank
[*] Looting /root/.ssh directory
[+] Downloaded /root/.ssh/authorized_keys -> ...
[*] Post module execution completed

The module no longer crashes. Instead, it prints a warning when credential storage fails and continues processing the remaining files. All SSH files are successfully downloaded, post-module execution completes normally, and the session remains active.

[4ravind-b]

@smcintyre-r7 the fix stops the crash and file downloads complete successfully, but credentials aren't saved to the DB in this scenario since the session record is gone. Is rescuing the exception and warning the user acceptable here, or would you prefer we attempt to save credentials to the current workspace as a fallback?