Assistance Disclosure

I consulted Gemini to help analyze the issue, locate the bug, and draft the fix and regression test. I have reviewed and verified that the code and logic are correct.

Fixes #7325.

What this PR does / why we need it:
In a previous PR (#5808), a nil pointer check was added for srv.AutoHTTPS to prevent a panic when evaluating srv.AutoHTTPS.Skip. However, because srv.AutoHTTPS is nil by default (when Auto HTTPS is fully enabled), this accidentally caused hasTLSEnabled to evaluate to false for site blocks without an explicit https:// scheme.

As a result, global TLS options like default_sni and fallback_sni were not propagated to tls_connection_policies during the adapter build phase, causing SNI-less clients to fail handshakes.

This PR fixes the logic by properly handling the nil state of srv.AutoHTTPS (i.e., treating nil as having no skip conditions):
srv.AutoHTTPS == nil || !slices.Contains(...)

I've also added a regression test to ensure default_sni is always properly attached to the generated JSON configuration, even without an explicit https:// block.