fix(Test): Fix table checksum test. Test on uncompressed.

[billprovince]

Problem

Fixes DGRAPHCORE-152

Previously, when testing to verify that a bad checksum would actually cause a checksum failure, we were overlooking the fact that our method of modifying the data could also modify the compressed data or even some of the metadata related to our block data.

The symptomatic results:

--- FAIL: TestTableChecksum (0.11s)
2640 table_test.go:698:
2641 Error Trace: table_test.go:698
2642 table_test.go:692
2643 Error: Received unexpected error:
2644 failed to initialize table error: failed to initialize biggest for table /tmp/2687635575.sst error: failed to decode compressed data in file: /tmp/2687635575.sst at offset: 51750, len: 149 error: failed to decompress err: unexpected EOF
2645 Test: TestTableChecksum

or some similar error

(There were earlier errors about failing to panic, but these were always generated before the secondary logged error was added).

Solution

Now, when we want to test that our checksum process is valid, we check with uncompressed data rather than with compressed data. In addition, we still write random bytes, but we no longer place it randomly. Instead, we place it in a location that will avoid the metadata locations of the data blocks.

[mangalaman93]

This looks good to me but I will let @harshil-goel or @joshua-goldstein approve it. Thanks for the PR Bill.

[joshua-goldstein]

Looks good, thanks Bill!