Significantly improve diffing performance and fix minor bug with bss section match percents

[LagoLunatic]

First commit optimizes diffing by moving the expensive string processing logic out of the loops being run to compare every symbol to every other symbol, and instead doing the logic just once while reading the object so that simple string equals checks can be used in the loop.

Second commit optimizes diffing more by using min in the loops instead of sorting.

Third commit optimizes this same part even further by removing the min in the loop and instead sorting all symbols once when the object is read. Moving the sorting earlier had the unintended side effect of fixing a subtle bug in diff_bss_section that I didn't know about: it assumes the bss symbols are in order and only compares their sizes, so when they're not in the right order it will report <100% if the bss symbol names don't match. Now it doesn't care about bss symbol names as long as the sizes match - this was the originally intended behavior for diff_bss_section, right?

[LagoLunatic]

Comparison for the bss section change:
Before:

After:

It now shows 100% even though the symbol names don't match. Seems like that was the intended behavior, though I'm not sure if that might be too lenient now?