"Location/range lists for this CU" is not a thing

[sevaa]

See this: #580

The way pyelftools handles location lists and rangelists, there is no way to properly support this kind of mixed bitness file structure. How loclist access normally goes, one constructs a LocationListParser around a LocationLists object, which comes from calling location_lists() on a DWARFInfo, which constructs a LocationLists with a copy of file level structs. No way to pass the structs of a specific CU there. Similar story with rangelists.

One can sort make this work by swapping the structs member on the LocationLists/RangeLists object before individual loclist/rangelist access, substituting the structs from the CU that holds the loclist/rangelist reference, but that's on the very edge of proper usage.

There is an optional DIE/CU parameter on get_location_list_at_offset() and similar; we could modify that to swap the structs to the current CU's structs if provided. This way the API surface would be intact (modulo the rather exotic method get_range_list_at_offset_ex(), and adding an extra optional argument is hardly a breaking change).

Alternatively, we could allow for explicitly constructing a loclist/rangelist parser for a specific CU. The former saves for memory but adds an extra check/set for every loclist access, the latter gives marginally better performance but can lead to proliferation of parser objects.