Axis titles depend on order of datasets

[ASmirnov-HORIS]

Look at the titles for y axis:

from lets_plot import *
LetsPlot.setup_html()

data1 = {'x': [0, 1], 'y': [0, 1]}
data2 = {'s': [1], 'i': [0]}

ggplot() + \
    geom_point(aes(x='x', y='y'), data=data1) + \
    geom_abline(aes(slope='s', intercept='i'), data=data2)

ggplot() + \
    geom_abline(aes(slope='s', intercept='i'), data=data2) + \
    geom_point(aes(x='x', y='y'), data=data1)

It seems that the axis title depends on the order of layers - each with its own data. If we do the same in ggplot2, in both cases axes will be titled "x" and "y".

[alshan]

It seems we should derive x/y axis labels from x/y aesthetics only and ignore other positional aesthetics.
If there are nothing to derive from, then just use "x", "y" labels.

See related #334

[OLarionova-HORIS]

PR: #342