Add a 'decode_timestamp' function

To help with decoding the nested tuple format which is how erlang's
pgo handles timestamps. Named as suggested by lpil.

I'm not sure the calling convention for the functions and so whether
or not it should be set up to use parens at the call site. I've opted
not to, to follow the convention of other value types like 'int' and
'string'.

src/gleam/pgo.gleam

@@ -493,3 +493,13 @@ pub fn error_code_name(error_code: String) -> Result(String, Nil) {
     _ -> Error(Nil)
   }
 }
+
+/// Checks to see if the value is formatted as `#(#(Int, Int, Int), #(Int, Int, Int))`
+/// to represent `#(#(year, month, day), #(hour, minute, second))`, and returns the
+/// value if it is.
+pub fn decode_timestamp(value: dynamic.Dynamic) {
+  dynamic.tuple2(
+    dynamic.tuple3(dynamic.int, dynamic.int, dynamic.int),
+    dynamic.tuple3(dynamic.int, dynamic.int, dynamic.int),
+  )(value)
+}

test/gleam/pgo_test.gleam

@@ -109,10 +109,7 @@ pub fn selecting_rows_test() {
         dynamic.int,
         dynamic.string,
         dynamic.bool,
-        dynamic.tuple2(
-          dynamic.tuple3(dynamic.int, dynamic.int, dynamic.int),
-          dynamic.tuple3(dynamic.int, dynamic.int, dynamic.int),
-        ),
+        pgo.decode_timestamp,
       ),
     )
 
@@ -319,10 +316,7 @@ pub fn datetime_test() {
     #(#(2022, 10, 10), #(11, 30, 30)),
     "timestamp",
     pgo.timestamp,
-    dynamic.tuple2(
-      dynamic.tuple3(dynamic.int, dynamic.int, dynamic.int),
-      dynamic.tuple3(dynamic.int, dynamic.int, dynamic.int),
-    ),
+    pgo.decode_timestamp,
   )
   |> pgo.disconnect
 }