adapt histograms to float/int

[OriolAbril]

Description

The align="left" we used as default in matplotlib distplot is great for ints, but super confusing
for floats (which are also valid histogram input via kind kwarg and/or rcparam). The plot
below shows the issue:

bokeh distplot used a mid/centered bins approach which is great for floats, but
requires remembering that intervals are [left_edge, right_edge) for ints. Now both
should behave in the same way, in both cases plotting histograms adapted to the dtype
of the input.

Checklist

• Does the PR follow official PR format?
• Is the code style correct (follows pylint and black guidelines)?
• Is the fix listed in the Maintenance and fixes section of the changelog?

---

📚 Documentation preview 📚: https://arviz--2247.org.readthedocs.build/en/2247/

[codecov]

Codecov Report

Merging #2247 (2b0a072) into main (61bda88) will increase coverage by 0.00%.
The diff coverage is 90.00%.

@@           Coverage Diff           @@
##             main    #2247   +/-   ##
=======================================
  Coverage   87.90%   87.91%           
=======================================
  Files         120      120           
  Lines       12417    12424    +7     
=======================================
+ Hits        10915    10922    +7     
  Misses       1502     1502           

Impacted Files	Coverage Δ
arviz/plots/backends/matplotlib/distplot.py	92.85% <87.50%> (+0.54%)	⬆️
arviz/plots/backends/bokeh/distplot.py	84.84% <100.00%> (+0.47%)	⬆️