boringssl: restore is_safe_alt_name helper (main build break)

[robobun]

What

Fixes the E0433 compile error on main since ff512ea.

Why

Merge-order race: #36252 removed the x509::is_safe_alt_name helper and its use x509 as X509 alias (the only caller had been rewritten), and #36165 then landed a new call to X509::is_safe_alt_name in the NameBytes Display impl that was written against the pre-#36252 tree. Every lane on build 86151 fails with:

error[E0433]: cannot find type `X509` in this scope
   --> src/boringssl/lib.rs:584:12
    |
584 |         if X509::is_safe_alt_name(self.0, false) {

Fix

Restore the helper as a private fn is_safe_alt_name(name: &[u8]) -> bool next to NameBytes. The sole caller passed utf8 = false, so the UTF-8 branch is dropped; the body is otherwise the same port of Node's IsSafeAltName (src/jsc/bindings/ncrypto.cpp:753) that #36252 removed.

Verification

• cargo check -p bun_boringssl and cargo clippy -p bun_boringssl --tests are clean
• cargo test -p bun_boringssl --lib runs the two new unit tests (safe_alt_name_matches_node, name_bytes_quoting)
• bun bd test test/js/node/tls/node-tls-connect-hostname-verification.test.ts -t "altname mismatch reason" passes all three escaping cases added in Robustness pass across install, css, ffi, crypto, spawn, shell, and node compat #36165

[claude]

LGTM — straightforward build-break fix restoring a removed helper.

What was reviewed:

• Verified is_safe_alt_name matches Node's IsSafeAltName non-UTF8 branch in ncrypto.cpp:753 byte-for-byte (range ' '..='~', four excluded chars).
• Confirmed no other callers of the removed X509::is_safe_alt_name remain in the tree.
• Unit tests cover the range boundaries (\x1f, \x7f, \xe4) and all four excluded characters.

Extended reasoning...

Overview

This PR fixes a compile error on main caused by a merge-order race between #36252 (removed x509::is_safe_alt_name) and #36165 (added a new caller in NameBytes::fmt). The fix restores the helper as a private free function next to its sole caller, drops the unused utf8 parameter branch, and adds two unit tests.

Security risks

None. The helper is a pure byte-classification predicate used only to decide whether a certificate SAN entry needs JSON-style quoting in an error message. It does not participate in hostname verification, trust decisions, or any parsing of untrusted input beyond display formatting. The implementation was cross-checked against src/jsc/bindings/ncrypto.cpp:753 and matches the non-UTF8 branch exactly: printable ASCII 0x20..=0x7E minus ", \\, ,, '.

Level of scrutiny

Low. This is a build-break fix with main currently red on every lane. The change is a 4-line function plus tests; the function body is a direct port of code that was already in-tree until #36252 removed it. The only judgment call — dropping the utf8 parameter — is justified: the sole caller passed false, and NameBytes is documented as IA5/Latin-1.

Other factors

• Grep confirms no other references to X509::is_safe_alt_name or the removed use x509 as X509 alias remain.
• The new unit tests exercise both range boundaries (\x1f below, \x7f above, non-ASCII \xe4) and every excluded character, plus the end-to-end NameBytes quoting output including the CVE-2021-44532-relevant comma-injection case.
• PR description cites cargo check/clippy/test and the integration test from #36165 all passing.