Xfail some conditions of signal tests under ROCm/HIP

[takagi]

Rel #4132, #4484.

Including fairly larger errors.

E       AssertionError: 
E       Not equal to tolerance rtol=0.001, atol=0.001
E       
E       Mismatched elements: 784 / 784 (100%)
E       Max absolute difference: 1333.3907
E       Max relative difference: 8.11097
E        x: array([[[ 1.132499e+02,  8.041714e+01,  1.015153e+02,  6.432927e+01,
E                 1.203019e+02,  1.354613e+01,  2.911143e+01, -2.381061e+01,
E                 2.635724e+01, -3.767127e+01,  1.562032e+01,  6.885651e+01,...
E        y: array([[[  30.119678,   78.50118 ,  117.3106  ,  146.02594 ,
E                 160.77292 ,  161.73344 ,  150.97159 ,  165.33815 ,
E                 196.18398 ,  194.92374 ,  152.63922 ,  113.40126 ,...

[takagi]

pfnCI, test this please.

[chainer-ci]

Jenkins CI test (for commit 6979c2a, target branch master) failed with status FAILURE.

[takagi]

pfnCI, test this please.

[chainer-ci]

Jenkins CI test (for commit 9ee6b69, target branch master) failed with status FAILURE.

[takagi]

pfnCI, test this please

[chainer-ci]

Jenkins CI test (for commit 9ee6b69, target branch master) failed with status FAILURE.

[takagi]

test_copyto_multigpu is known to be fleaky and is not relevant to this change.

[emcastillo]

Probably all these tests have issues related to memory fences/warp sizes etc.

[emcastillo]

Jenkins, test this please

[chainer-ci]

Jenkins CI test (for commit 9ee6b69, target branch master) succeeded!