Add relplot function

[mwaskom]

This PR adds the relplot function, for combining scatterplot or lineplot with a FacetGrid, similar to the relationship between factorplot and the axes-level categorical plots. This makes it easy and robust to facet relational plots with multiple semantic variables:

diamonds = sns.load_dataset("diamonds")
clarity_ranking = ["I1", "SI2", "SI1", "VS2", "VS1", "VVS2", "VVS1", "IF"]
g = sns.relplot(x="carat", y="price",
                hue="clarity", size="depth", col="color",
                height=3.5, aspect=.9, col_wrap=3,
                palette="ch:.2,-.2,d=.3_r",
                hue_order=clarity_ranking,
                linewidth=0, sizes=(5, 20),
                data=diamonds.sort_values("color"))

fmri = sns.load_dataset("fmri")
sns.relplot(x="timepoint", y="signal",
            hue="event", style="event",
            col="region", units="subject",
            kind="line", estimator=None,
            height=5, aspect=.8, data=fmri)

This addition necessitated an API change in FacetGrid, because there was a conflict with the size parameter name. This prompted a uniform substitution of height for size across grid functions, which has the benefit of being more explicit about what that parameter does. It also fixes a problem in factorplot where there was an existing conflict where size was used for the marker radius. The change should not break existing code, as size is handled and converted set to height. with a warning issued.

[codecov]

Codecov Report

Merging #1477 into master will decrease coverage by 0.31%.
The diff coverage is 76.92%.

@@            Coverage Diff             @@
##           master    #1477      +/-   ##
==========================================
- Coverage   94.51%   94.19%   -0.32%     
==========================================
  Files          15       15              
  Lines        4411     4479      +68     
==========================================
+ Hits         4169     4219      +50     
- Misses        242      260      +18

Impacted Files	Coverage Δ
seaborn/categorical.py	96.69% <100%> (+0.01%)	⬆️
seaborn/basic.py	99.45% <100%> (+0.04%)	⬆️
seaborn/regression.py	97.01% <40%> (-0.96%)	⬇️
seaborn/axisgrid.py	94.72% <48.27%> (-1.84%)	⬇️

---

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 56743db...62fb1ee. Read the comment docs.