Sema: fix analyzeBlockBody logic

Previously, when a coercion needed to be inserted into a break
instruction, the `br` AIR instruction would be rewritten so that the
block operand was a sub-block that did the coercion. The problem is that
the sub-block itself was never added to the parent block, resulting in
the `br` instruction operand being a bad reference.

Now, the `br` AIR instruction that needs to have coercion instructions
added is replaced with the sub-block itself with type `noreturn`, and
then the sub-block has the coercion instructions and a new `br`
instruction that breaks from the original block.

LLVM backend needed to be fixed to lower `noreturn` blocks without
emitting an unused LLVM basic block.

src/codegen/llvm.zig

@@ -2063,8 +2063,14 @@ pub const FuncGen = struct {
         const ty_pl = self.air.instructions.items(.data)[inst].ty_pl;
         const extra = self.air.extraData(Air.Block, ty_pl.payload);
         const body = self.air.extra[extra.end..][0..extra.data.body_len];
+        const inst_ty = self.air.typeOfIndex(inst);
         const parent_bb = self.context.createBasicBlock("Block");
 
+        if (inst_ty.isNoReturn()) {
+            try self.genBody(body);
+            return null;
+        }
+
         var break_bbs: BreakBasicBlocks = .{};
         defer break_bbs.deinit(self.gpa);
 
@@ -2084,7 +2090,6 @@ pub const FuncGen = struct {
         self.builder.positionBuilderAtEnd(parent_bb);
 
         // If the block does not return a value, we dont have to create a phi node.
-        const inst_ty = self.air.typeOfIndex(inst);
         if (!inst_ty.hasCodeGenBits()) return null;
 
         const raw_llvm_ty = try self.dg.llvmType(inst_ty);