Rethink sorting fixes by start

[iisaduan]

Currently, the tool accepts all fixes that don't collide, and if there are colliding fixes, the tool then picks the fix that occurs first by start index. This however, can produce several unwanted effects in the resulting code, such as applying multiple fixes for one error, or prioritizing changes by how the code is organized, instead of which fix is most likely to be applied. As written in #5, this is already done by services, and could be used to improve the output of the tool.