Don't crash on non-UTF-8 bytes in the map file

[munzzyy]

Fixes #2.

parse_sections() opens the map file with the platform default encoding and no error handling, so one non-UTF-8 byte anywhere in the file kills the parse. Linker maps quote build paths verbatim, so this happens as soon as a path contains a character that isn't ASCII — which is what the reporter in #2 hit on Fedora.

Reproduced with a map file that is otherwise well formed, with a single 0xe9 byte inside a path in a .text entry:

  File "map_parser.py", line 62, in parse_sections
    line = file.readline()
  File "<frozen codecs>", line 325, in decode
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xe9 in position 243: invalid continuation byte

Reading with errors='replace' parses the same file cleanly. Only the offending byte in the path degrades to U+FFFD; every address and size in the output is unchanged, since those are plain ASCII hex on their own columns.

Second, smaller thing in the same function: the "memory configuration not found" path raises using input_file, which isn't a parameter of parse_sections — it only exists as a module-level name when the file is run as a script. Import the function and call it directly on a map with no Memory Configuration header and you get NameError: name 'input_file' is not defined instead of the intended message. Switched it to the function's own file_name argument, and dropped the missing space after "the".