Fix convolution spec and outputs in README.md

Closes #35 Updated the conv_spec function in the .py and .ipynb files. Updated README.md to match the changes. This version of conv_spec was confirmed with the test case inputs against `numpy.convolution(a, b, mode='same')` and `cupy.convolution(a, b, mode='same')`. All 3 calculations are in agreement for both test cases. The spec change doesn't make the problem particularly harder. The solution would require substituting `i - j + 1` in place of `i + j` and adding a check for `i - j + 1 > size`. Due to the added check, it may be worth updating the `conv_test` text from `# FILL ME IN (roughly 17 lines)` to `# FILL ME IN (roughly 18 lines)`, however it's only a rough number of lines regardless.