The current differ assumes the line index always starts at zero and the

get_opcodes() function in MyersDiffer calculates the line number base

on this assumption.
However, one future differ option, PatienceDiffer, which uses

MyersDiffer to obtain the final diff, passes in portions of the overall

file seperately and the start index will not be line 0 for code chunks

that do not start at the head of the file. As a result, some index

varables needs to be set in the differ to keep track of which line

each code block starts at.
This commit adds such two new parameters, a_start_index and

b_start_index in the base differ class and updates MyersDiffer to

generate correct opcode after the new change.

Four new test cases are added to make sure MyersDiffer works, given

different start indices for file a and b, for all the opertaions: equal,

replace, insert and delete. Also, another test case with same non-zero

start indices is added to make sure MyersDiffer works when the code

chunk is at the same location but not the start of the file.