saddle-point: use unambiguous coordinates

[bitfield]

The description for 'Saddle Point' illustrates what a saddle point is by showing an example matrix:

    0  1  2
  |---------
0 | 9  8  7
1 | 5  3  2     <--- saddle point at (1,0)
2 | 6  6  7

Because there are at least two completely reasonable coordinate systems which could apply here, (1,0) is ambiguous (does it mean "column 0, row 1", which is in fact correct, or "row 1, column 0"?)

This PR makes it unambiguous:

    0  1  2
  |---------
0 | 9  8  7
1 | 5  3  2     <--- saddle point at column 0, row 1, with value 5
2 | 6  6  7

Fixes #313.

This is also a good complement to the test data, which specifies:

      "expected": [
        {
          "row": 1,
          "column": 0
        }

[rpottsoh]

What I see looks fine but I am wondering if everything addressed in #313 is fixed here. This comment stuck out to me when I quickly scanned #313 just now. If you decide to pursue re-basing the coordinate system to be 1 based the changes will spill over into the canonical data.

[rpottsoh]

Matrix was recently updated to follow a more canonical notation for matrix indices. See #1387.

[bitfield]

Fine. As far I can see, the issue of coordinate base is orthogonal to the issue of coordinate ordering (that's why I didn't address it in this PR). I'm happy to update it to follow the Matrix example.

[rpottsoh]

LGTM. This probably doesn't need to be squashed when merged. Thanks for working on this @bitfield

[petertseng]

[rpottsoh]

Nice catch. Thanks

[petertseng]

#1433