Added decoding for base64-encoded images

[jbednar]

Right now, Image types accept a binary bytes object and return a base64-encoded HTML representation of the image suitable for embedding in an HTML document. Some packages like Datashader return such a bytes object from their _repr_png_() methods, which works well, but other _repr_png_ methods like the ones in nxpd return a string already base64 encoded. With this PR, if a string is provided as the image data, Panel will base64-decode the string before moving on; it will later be base64 encoded again. This double decoding and subsequent re-encoding isn't optimal, but it's a simple fix for not being able to view such objects.

The code uses isinstance(...,str) to check for the string, which seems safe for py2/py3 because a base64-encoded string should not normally be unicode on either Python version.

[codecov-io]

Codecov Report

Merging #88 into master will decrease coverage by 0.02%.
The diff coverage is 50%.

@@            Coverage Diff             @@
##           master      #88      +/-   ##
==========================================
- Coverage   91.69%   91.67%   -0.03%     
==========================================
  Files          23       23              
  Lines        3674     3676       +2     
==========================================
+ Hits         3369     3370       +1     
- Misses        305      306       +1

Impacted Files	Coverage Δ
panel/pane.py	81.36% <50%> (-0.16%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1547390...44452ba. Read the comment docs.