added support for shrink axis mask for StridedSlice op

[pyu10055]

Description

This is required for the SSD model, which has StridedSlice op with shrinkAxisMask set.

---

For repository owners only:

Please remember to apply all applicable tags to your pull request.
Tags: FEATURE, BREAKING, BUG, PERF, DEV, DOC, SECURITY

For more info see: https://github.com/tensorflow/tfjs/blob/master/DEVELOPMENT.md

---

This change is 

[dsmilkov]

Reviewed 5 of 7 files at r1.
Reviewable status: 0 of 1 approvals obtained (waiting on @pyu10055 and @dsmilkov)

---

src/ops/slice_util.ts, line 57 at r1 (raw file):

    startIndex[i] = startForAxis(beginMask, begin, strides, shape, i);
    endIndex[i] = stopForAxis(endMask, end, strides, shape, i);
    // When shrinking an axis, user startIndex + 1 for endIndex.

typo: user => use ?

---

src/ops/strided_slice.ts, line 58 at r1 (raw file):

function stridedSlice_<T extends Tensor>(
    x: T|TensorLike, begin: number[], end: number[], strides: number[],
    beginMask = 0, endMask = 0, shrinkAxisMask = 0): T {

To protect from breaking the API in the future when we add support for the other masks, add ellipsisMask and newAxisMask before shrinkAxisMarkand default them to 0. Then throw an error if their values are other than zero saying 'newAxisMark != 0 is not yet supported'. Likewise for elipsisMask

---

src/ops/strided_slice_test.ts, line 26 at r1 (raw file):

    const tensor = tf.tensor1d([0, 1, 2, 3]);
    const output = tf.stridedSlice(tensor, [0], [3], [2]);
    expect(output.shape).toEqual([2]);

the merge is strange here, since you dropped 3 unit tests

[dsmilkov]

1 tiny comment - add the elipsisMask and newAxisMask to the backend.ts as well so we can easily enable those for the nodejs backend later