Trim down/split up setup section of notebooks tutorial

The “TensorBoard in notebooks” tutorial starts with setup
instructions, which makes sense. But while the setup instructions for
90% of users are just %load_ext tensorboard, the first four paragraphs
of the section, including the first code block, are for niche edge cases
like cross-virtualenv Jupyter installs and Docker users. This makes it
hard to find what most users should actually run.

We should consider relegating the edge-case setup instructions to the
end of the notebook, and put a note referring to them after the common
setup:

# Setup

Start by installing TF 2.0 and loading the TensorBoard notebook extension:

```python
%load_ext tensorboard
```

**For Jupyter users:** If you use a non-standard virtualenv setup and
are running into problems, see “Advanced setup” below.

**For Docker users:** See “Advanced setup” below.

# TensorBoard in notebooks

Download the [FashionMNIST] dataset and scale it:

```python
import datetime, os
```

```python
fashion_mnist = ...
```

Open the linked “TensorBoard in notebooks” tutorial and inspect its Setup section and first code block. Move the cross-virtualenv Jupyter and Docker guidance into an Advanced setup section after the common %load_ext tensorboard instruction, with notes linking to it; done when the common setup is immediately visible and the edge cases remain documented.