[feat] Parse convertible mapping objects with json module

The advantage is that we can automatically convert nested mappings, like for
example the `extra_resources` attribute, which can thus be set from the command
line.

reframe/utility/typecheck.py

@@ -95,6 +95,7 @@
 
 import abc
 import datetime
+import json
 import re
 
 
@@ -320,21 +321,12 @@ def __getitem__(cls, typespec):
 
     def __rfm_cast_str__(cls, s):
         mappping_type = cls._type
-        key_type = cls._key_type
-        value_type = cls._value_type
-        seq = []
-        for key_datum in s.split(','):
-            try:
-                k, v = key_datum.split(':')
-            except ValueError:
-                # Re-raise as TypeError
-                raise TypeError(
-                    f'cannot convert string {s!r} to {cls.__name__!r}'
-                ) from None
-
-            seq.append((key_type(k), value_type(v)))
-
-        return mappping_type(seq)
+        # Convert string to JSON format: wrap it with braces and add double
+        # quotes where necessary.
+        s = f'{{{s}}}'
+        s = re.sub('([:,{] *)([^":,{}]+)', r'\1"\2"', s)
+        # Parse the JSON string and convert to the desired mapping type.
+        return mappping_type(json.loads(s))
 
 
 class _StrType(_SequenceType):