webgl: Add packed op for sin/cos

[qjia7]

Fixed #7372

To see the logs from the Cloud Build CI, please join either our discussion or announcement mailing list.

---

This change is 

[qjia7]

Based on below profiling result. Many time are spent on PackProgram and UnpackProgram. The reason is that we didn't provide a packed Sin op. So the sequence will be like UnpackProgram -> Sin -> PackProgram -> MatMulPackedProgram, which brings many extra overhead. With this change, the time of the user provided model will be reduced by more than a third (712 ms ->449 ms) in my machine.

Kernel	Time(ms)	Inputs	Output	GPUPrograms
_FusedMatMul	115.86	input0: 3D[1,262144,256]input1: 3D[1,256,256]input2: 1D[256]input3: null	1,262144,256	PackProgram: 34.802812, MatMulPackedProgram: 81.055208
_FusedMatMul	112.14	input0: 3D[1,262144,256]input1: 3D[1,256,256]input2: 1D[256]input3: null	1,262144,256	PackProgram: 33.896666, MatMulPackedProgram: 78.238645
_FusedMatMul	111.30	input0: 3D[1,262144,256]input1: 3D[1,256,256]input2: 1D[256]input3: null	1,262144,256	PackProgram: 34.840781, MatMulPackedProgram: 76.462864
_FusedMatMul	65.40	input0: 3D[1,262144,256]input1: 3D[1,256,3]input2: 1D[3]input3: null	1,262144,3	PackProgram: 34.896979, MatMulPackedProgram: 30.505208
Sin	57.77	input0: 2D[262144,256]	262144,256	UnpackProgram: 29.14401, UnaryOpProgram: 28.624062
Sin	56.38	input0: 2D[262144,256]	262144,256	UnpackProgram: 28.064687, UnaryOpProgram: 28.311875
Sin	55.10	input0: 2D[262144,256]	262144,256	UnpackProgram: 27.104322, UnaryOpProgram: 27.99526
Sin	54.84	input0: 2D[262144,256]	262144,256	UnpackProgram: 28.026562, UnaryOpProgram: 26.813541

[pyu10055]

Reviewed 2 of 2 files at r1, all commit messages.
Reviewable status: complete! 1 of 1 approvals obtained (waiting on @Linchenn)

[Linchenn]

Thank you for fixing it! LGTM