Add popup dialog for plugins that require TensorFlow

[orionr]

Adding a popup dialog for when plugins are not created. Note that this is different from being "inactive", which can be due to data, etc.

Fixes #2027

Everything should be good to go and I've tested on my Mac with the following:

conda create --name tfpy36 python=3.6
conda activate tfpy36
pip install --upgrade pip
pip install tf-nightly-2.0-preview
conda install protobuf
conda install absl-py
conda install numpy
bazel build tensorboard
./bazel-bin/tensorboard/tensorboard --logdir ~/local/sample_runs/runs
# Navigate to the hparams tab and verify a popup is NOT shown

conda create --name notf python=2.7
conda activate notf
pip install --upgrade pip
pip install scipy
conda install protobuf
conda install absl-py
conda install numpy
bazel build tensorboard
./bazel-bin/tensorboard/tensorboard --logdir ~/local/sample_runs/runs
# Navigate to the hparams tab and verify a popup is shown

cc @lanpa, @nfelt

[orionr]

@nfelt if tests pass this should be good for review. Thanks!

[orionr]

And tests pass! We're good.

[orionr]

@nfelt and @wchargin, a gentle ping for review on this. Thanks.

[nfelt]

Looks good overall, just a few pretty minor code cleanup requests.

The debugger plugin currently doesn't have any indication about it not working with TF - can we change the recently introduced try-catch in the loader to raise an exception indicating that TF is required to use the debugger? This only happens when the --debugger_port has been passed so it doesn't impact the default no-TF use case.

tensorboard/tensorboard/plugins/debugger/debugger_plugin_loader.py

Lines 98 to 102 in 90a386d

	try:
	# pylint: disable=g-import-not-at-top,unused-import
	import tensorflow
	except ImportError:
	return None

Ideally we would message that through the dialog, but the way the debugger is structured with its own separate initial dialog makes that difficult, so for now this would at least prevent if from silently not working.

[orionr]

@nfelt (and @wchargin) all comments should be addressed. I needed to use a function scope to add PLUGIN_NAME but that seemed worth it. Also kept the _requestManager change, since it is consistent with the other plugins. Let me know if I should revert that particular change. Tested in a non-TensorFlow and TensorFlow environment locally on my Mac.

[orionr]

@nfelt any thoughts on the CI failures? Seems unrelated to my changes, but not 100% sure.

[nfelt]

Thanks for the PLUGIN_NAME cleanup!

[orionr]

@nfelt ready for final review. Thanks.

[orionr]

One last else clause coming - stay tuned.

[orionr]

@nfelt made that one last fix - thanks.

[orionr]

@nfelt, please merge when ready. Thank you. :)