Refactor legend+colorbar internals, support "queueing" and location-based replacements

[lukelbd]

Resolves #198. Proplot supports more than one legend per subplot. However this means that each legend() call creates a brand new legend -- sometimes one on top of another. This was confusing to new users (IIRC there's a closed issue floating around) and broke external packages that rely on one legend per subplot (#198).

This PR does the best of both worlds. Proplot still supports multiple legends, but not on top of each other. If you call legend() more than once with the same location loc (or without loc, implying the same default location), proplot
will either 1) replace the previous legend (if queue=False, the default) or 2) add handles to a "queue" (if queue=True). This "queueing" option was previously only available in the wrapper functions, which didn't make sense.

This PR also cleans up/organizes the messy legend and colorbar code. It also changes behavior: legend(..., center=True) no longer returns the background patch. I think it's kind of weird for legend() to return a non-legend object and "inner colorbars" made with e.g. colorbar(..., loc='upper right') also don't return the patch. Really should just merge #74.

Here's the example from #198 (no more errors):

import pandas as pd
import proplot as plot
df = pd.DataFrame({'A' : ['foo', 'bar', 'foo', 'bar',
                          'foo', 'bar', 'foo', 'foo'],
                   'B' : ['one', 'one', 'two', 'three',
                       'two', 'two', 'one', 'three'],
                   'C' : np.random.randn(8),
                   'D' : np.random.randn(8)})
fig, axs = plot.subplots()
df.groupby('A')['C'].plot(legend='ur', ax=axs)

Here's an example showing the new location-specific legend-replacing behavior:

import proplot as plot
data = np.random.rand(10, 4)
fig, ax = plot.subplots()
hs = ax.plot(data[:, :2], labels=('foo', 'bar'))
ax.legend(hs, loc='ur')
hs = ax.plot(data[:, 2:], labels=('abc', 'def'))  # overwrites old one
ax.legend(hs, loc='ur')
ax.legend(hs, loc='top')

And here's an example of the now publicly-available queue feature:

import proplot as plot
data = np.random.rand(10, 4)
fig, ax = plot.subplots()
hs = ax.plot(data[:, 0], label='foo')
ax.legend(hs,  queue=True)
hs = ax.plot(data[:, 1], label='bar')
ax.legend(hs,  queue=True)
hs = ax.plot(data[:, 2], label='abc')
ax.legend(hs, loc='b', queue=True)
hs = ax.plot(data[:, 3], label='def')
ax.legend(hs, loc='b', queue=True)