x86_64: implement more shuffles

Author: jacobly0

lib/std/crypto/blake3.zig

@@ -200,8 +200,7 @@ const CompressGeneric = struct {
     }
 };
 
-const compress = if (builtin.cpu.arch == .x86_64 and
-    (builtin.zig_backend != .stage2_x86_64 or std.Target.x86.featureSetHas(builtin.cpu.features, .ssse3)))
+const compress = if (builtin.cpu.arch == .x86_64)
     CompressVectorized.compress
 else
     CompressGeneric.compress;

lib/std/crypto/salsa20.zig

@@ -302,8 +302,7 @@ fn SalsaNonVecImpl(comptime rounds: comptime_int) type {
     };
 }
 
-const SalsaImpl = if (builtin.cpu.arch == .x86_64 and
-    (builtin.zig_backend != .stage2_x86_64 or std.Target.x86.featureSetHas(builtin.cpu.features, .ssse3)))
+const SalsaImpl = if (builtin.cpu.arch == .x86_64)
     SalsaVecImpl
 else
     SalsaNonVecImpl;