compiler: add support for running a builtin in a goroutine

Not sure why you would ever do this, but it appears to be allowed by the
Go specification and previously TinyGo would crash with an unhelpful
error message when you would do this. I don't see any practical use of
it.

The implementation simply runs the builtin directly.

compiler/testdata/goroutine-cortex-m-qemu.ll

@@ -3,6 +3,12 @@ source_filename = "goroutine.go"
 target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
 target triple = "armv7m-none-eabi"
 
+%runtime.channel = type { i32, i32, i8, %runtime.channelBlockedList*, i32, i32, i32, i8* }
+%runtime.channelBlockedList = type { %runtime.channelBlockedList*, %"internal/task.Task"*, %runtime.chanSelectState*, { %runtime.channelBlockedList*, i32, i32 } }
+%"internal/task.Task" = type { %"internal/task.Task"*, i8*, i32, %"internal/task.state" }
+%"internal/task.state" = type { i32, i32* }
+%runtime.chanSelectState = type { %runtime.channel*, i8* }
+
 @"main.regularFunctionGoroutine$pack" = private unnamed_addr constant { i32, i8* } { i32 5, i8* undef }
 @"main.inlineFunctionGoroutine$pack" = private unnamed_addr constant { i32, i8* } { i32 5, i8* undef }
 
@@ -138,6 +144,27 @@ entry:
   ret void
 }
 
+define hidden void @main.recoverBuiltinGoroutine(i8* %context, i8* %parentHandle) unnamed_addr {
+entry:
+  ret void
+}
+
+define hidden void @main.copyBuiltinGoroutine(i8* %dst.data, i32 %dst.len, i32 %dst.cap, i8* %src.data, i32 %src.len, i32 %src.cap, i8* %context, i8* %parentHandle) unnamed_addr {
+entry:
+  %copy.n = call i32 @runtime.sliceCopy(i8* %dst.data, i8* %src.data, i32 %dst.len, i32 %src.len, i32 1, i8* undef, i8* null)
+  ret void
+}
+
+declare i32 @runtime.sliceCopy(i8* nocapture writeonly, i8* nocapture readonly, i32, i32, i32, i8*, i8*)
+
+define hidden void @main.closeBuiltinGoroutine(%runtime.channel* dereferenceable_or_null(32) %ch, i8* %context, i8* %parentHandle) unnamed_addr {
+entry:
+  call void @runtime.chanClose(%runtime.channel* %ch, i8* undef, i8* null)
+  ret void
+}
+
+declare void @runtime.chanClose(%runtime.channel* dereferenceable_or_null(32), i8*, i8*)
+
 attributes #0 = { "tinygo-gowrapper"="main.regularFunction" }
 attributes #1 = { "tinygo-gowrapper"="main.inlineFunctionGoroutine$1" }
 attributes #2 = { "tinygo-gowrapper"="main.closureFunctionGoroutine$1" }

compiler/testdata/goroutine-wasm.ll

@@ -4,6 +4,11 @@ target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
 target triple = "wasm32--wasi"
 
 %runtime.funcValueWithSignature = type { i32, i8* }
+%runtime.channel = type { i32, i32, i8, %runtime.channelBlockedList*, i32, i32, i32, i8* }
+%runtime.channelBlockedList = type { %runtime.channelBlockedList*, %"internal/task.Task"*, %runtime.chanSelectState*, { %runtime.channelBlockedList*, i32, i32 } }
+%"internal/task.Task" = type { %"internal/task.Task"*, i8*, i32, %"internal/task.state" }
+%"internal/task.state" = type { i8* }
+%runtime.chanSelectState = type { %runtime.channel*, i8* }
 
 @"main.regularFunctionGoroutine$pack" = private unnamed_addr constant { i32, i8* } { i32 5, i8* undef }
 @"main.inlineFunctionGoroutine$pack" = private unnamed_addr constant { i32, i8* } { i32 5, i8* undef }
@@ -88,3 +93,24 @@ entry:
 }
 
 declare i32 @runtime.getFuncPtr(i8*, i32, i8* dereferenceable_or_null(1), i8*, i8*)
+
+define hidden void @main.recoverBuiltinGoroutine(i8* %context, i8* %parentHandle) unnamed_addr {
+entry:
+  ret void
+}
+
+define hidden void @main.copyBuiltinGoroutine(i8* %dst.data, i32 %dst.len, i32 %dst.cap, i8* %src.data, i32 %src.len, i32 %src.cap, i8* %context, i8* %parentHandle) unnamed_addr {
+entry:
+  %copy.n = call i32 @runtime.sliceCopy(i8* %dst.data, i8* %src.data, i32 %dst.len, i32 %src.len, i32 1, i8* undef, i8* null)
+  ret void
+}
+
+declare i32 @runtime.sliceCopy(i8* nocapture writeonly, i8* nocapture readonly, i32, i32, i32, i8*, i8*)
+
+define hidden void @main.closeBuiltinGoroutine(%runtime.channel* dereferenceable_or_null(32) %ch, i8* %context, i8* %parentHandle) unnamed_addr {
+entry:
+  call void @runtime.chanClose(%runtime.channel* %ch, i8* undef, i8* null)
+  ret void
+}
+
+declare void @runtime.chanClose(%runtime.channel* dereferenceable_or_null(32), i8*, i8*)

compiler/testdata/goroutine.go

@@ -21,4 +21,23 @@ func funcGoroutine(fn func(x int)) {
 	go fn(5)
 }
 
+func recoverBuiltinGoroutine() {
+	// This is a no-op.
+	go recover()
+}
+
+func copyBuiltinGoroutine(dst, src []byte) {
+	// This is not run in a goroutine. While this copy operation can indeed take
+	// some time (if there is a lot of data to copy), there is no race-free way
+	// to make use of the result so it's unlikely applications will make use of
+	// it. And doing it this way should be just within the Go specification.
+	go copy(dst, src)
+}
+
+func closeBuiltinGoroutine(ch chan int) {
+	// This builtin is executed directly, not in a goroutine.
+	// The observed behavior is the same.
+	go close(ch)
+}
+
 func regularFunction(x int)