Only set complevel for compressed float vars

[zmoon]

We first noticed this in #160

In current netcdf-c (v4.9.1 when first noticed), it errors if you set compression for a variable-length dtype variable (such as string arrays with NumPy object dtype). With this PR, we only set compression for float vars.

See:

• Explicitly disallow variable length type compression Unidata/netcdf-c#2231
• Filter error: bad id or parameters or duplicate filter Unidata/netcdf4-python#1205
• Filters applied to variable length data Unidata/netcdf-c#2554

---

Summary:

• in write_ncf, only supply complevel for float vars
• add save/load to idealized example nb, which gets run in CI
• only supply complevel for float vars in write_analysis_ncf
• add some links/markup to the separate save/load example nbs
• fix toggle buttons in some example notebooks by explicitly adding sphinx-togglebutton

[zmoon]

@colin-harkins we may have to do this in write_analysis_ncf as well. Maybe we can add a test. ✔

[zmoon]

@colin-harkins I think this is ready for you to take a look. I added some links and such to your example nbs.

[colin-harkins]

@zmoon This looks good to me, thanks!