Localize/convert transformed_data datetimes to local_tz

[jonmmease]

Previously datetime columns returned by vf.transformed_data were naive but implicitly in the UTC timezone. This PR localizes them as UTC then converts them to VegaFusion's local timezone.

For example:

import vegafusion as vf
import altair as alt
from vega_datasets import data

# Manually set timezone to Seattle's since this a seattle weather
# dataset
vf.set_local_tz("America/Los_Angeles")

source = data.seattle_weather()

chart = alt.Chart(source).mark_bar(
    cornerRadiusTopLeft=3,
    cornerRadiusTopRight=3
).encode(
    x='month(date):O',
    y='count():Q',
    color='weather:N'
)
chart

tx_df = vf.transformed_data(chart, row_limit=5)
tx_df

	weather	month_date	__count	__count_start	__count_end
0	drizzle	2012-01-01 00:00:00-08:00	10	114	124
1	rain	2012-01-01 00:00:00-08:00	35	41	76
2	sun	2012-01-01 00:00:00-08:00	33	0	33
3	snow	2012-01-01 00:00:00-08:00	8	33	41
4	rain	2012-02-01 00:00:00-08:00	40	33	73

tx_df.dtypes

weather                                       object
month_date       datetime64[ns, America/Los_Angeles]
__count                                        int64
__count_start                                  int64
__count_end                                    int64
dtype: object