Guard against word-not-found in sread tokenization

If text.find(word_str, offset) returns -1, use the offset as a fallback.

src/mudlib/zmachine/zcpu.py

@@ -670,6 +670,8 @@ class ZCpu:
                 word_str, dict_addr = tokens[i]
                 # Find word position in text
                 pos = text.find(word_str, offset)
+                if pos == -1:
+                    pos = offset  # fallback: best guess position
                 word_len = len(word_str)
                 base = parse_buffer_addr + 2 + (i * 4)
                 self._memory.write_word(base, dict_addr)