mx.random.seed with ctx failures on a gpu build when run with cpu context

[anirudh2290]

Minimum reproducible example.

import mxnet as mx

def set_seed_variously_for_context(ctx, init_seed, num_init_seeds, final_seed):
    end_seed = init_seed + num_init_seeds
    for seed in range(init_seed, end_seed):
        mx.random.seed(seed, ctx=ctx)
    z = mx.random.seed(seed, ctx=ctx)
    return end_seed

dtype = 'float32'

samples_imp = []
samples_sym = []
ctx = mx.cpu()
shape = (200, 200)
params = {'low': -1.5, 'high': 3.0}
params.update(shape=shape, dtype=dtype)
seed = set_seed_variously_for_context(ctx, 1, 1, 1234)
mx.nd.waitall()

To reproduce, this will require the exception handling support for waitall in this PR: #14397 . This issue was found because of CI failures when running test_random.py on windows. It was hidden earlier because waitall didnt have exception rethrow support. This issue may have been around since the PR was added: #10367

Currently working on fixing this.

[mxnet-label-bot]

Hey, this is the MXNet Label Bot.
Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it.
Here are my recommended labels: Bug