Add single <-> double conversions.

- With explicit casting now possible, converting from double to single
  precision floating point and back again needs to be possible, too.

src/zscript/ast.cpp

@@ -160,10 +160,15 @@ public:
 		size_t len = mysnprintf(buf, countof(buf), "%08x", x);
 		Add(buf, len);
 	}
-	void AddFloat(double f)
+	void AddFloat(double f, bool single)
 	{
 		char buf[32];
 		size_t len = mysnprintf(buf, countof(buf), "%.4f", f);
+		if (single)
+		{
+			buf[len++] = 'f';
+			buf[len] = '\0';
+		}
 		Add(buf, len);
 	}
 private:
@@ -501,7 +506,11 @@ static void PrintExprConstant(FLispString &out, ZCC_TreeNode *node)
 	}
 	else if (enode->Type == TypeFloat64)
 	{
-		out.AddFloat(enode->DoubleVal);
+		out.AddFloat(enode->DoubleVal, false);
+	}
+	else if (enode->Type == TypeFloat32)
+	{
+		out.AddFloat(enode->DoubleVal, true);
 	}
 	else if (enode->Type == TypeName)
 	{

src/zscript/zcc_expr.cpp

@@ -613,6 +613,21 @@ static void DtoU32(ZCC_ExprConstant *expr, FSharedStringArena &str_arena)
 	expr->Type = TypeUInt32;
 }
 
+static void FtoD(ZCC_ExprConstant *expr, FSharedStringArena &str_arena)
+{
+	// Constant single precision numbers are stored as doubles.
+	assert(expr->Type == TypeFloat32);
+	expr->Type = TypeFloat64;
+}
+
+static void DtoF(ZCC_ExprConstant *expr, FSharedStringArena &str_arena)
+{
+	// Truncate double precision to single precision.
+	float poop = (float)expr->DoubleVal;
+	expr->DoubleVal = poop;
+	expr->Type = TypeFloat32;
+}
+
 static void S32toS(ZCC_ExprConstant *expr, FSharedStringArena &str_arena)
 {
 	char str[16];
@@ -664,7 +679,10 @@ void ZCC_InitConversions()
 	TypeSInt32->AddConversion(TypeFloat64, S32toD);
 	TypeSInt32->AddConversion(TypeString, S32toS);
 
+	TypeFloat32->AddConversion(TypeFloat64, FtoD);
+
 	TypeFloat64->AddConversion(TypeUInt32, DtoU32);
 	TypeFloat64->AddConversion(TypeSInt32, DtoS32);
+	TypeFloat64->AddConversion(TypeFloat32, DtoF);
 	TypeFloat64->AddConversion(TypeString, DtoS);
 }