Update dynamically computed signature

[maximlt]

The signature of each plotting method is dynamically constructed on import. They include:

1. Original method signature (e.g. of .line()) without variadic keywords
2. Valid backend style options for the element (obtained _kind_mapping) and backend (defaults to bokeh)
3. From the mapping HoloViewsConverter._kind_options (plotting method name to a short list of options)
4. From the mapping HoloViewsConverter._axis_options (list of axis options)
5. From the mapping HoloViewsConverter._op_options (list of operation options)
6. **kwargs

2), 3), 4) and 5) are only built as inspect.Parameter(k, inspect.Parameter.KEYWORD_ONLY) so they don't include the default values of these options, that could, for example, be defined in HoloViewsConverter.__init__.

Many options are in fact missing, here's the list obtained by comparing the signature of HoloViewsConverter.__init__ with hvplot.plotting.hvTabular.line:

attr_labels
autorange
backlog
by
check_symmetric_max
clim
cnorm
coastline
col
crs
cticks
debug
dynamic
dynspread
features
fields
flip_xaxis
flip_yaxis
framewise
geo
global_extent
group_label
groupby
hover
hover_cols
hover_formatters
hover_tooltips
kind
kwds
label
persist
precompute
project
projection
rescale_discrete_levels
robust
row
sort_date
stacked
stream
subplots
symmetric
tiles
tiles_opts
tools
transforms
use_dask
use_index
value_label

Most of these options should be added to the signature with their default value (not all, e.g. kind).

• Add missing options to the method signature

Despite these missing options, the signature of .line() includes 95 members. 47 of them are style options (e.g. hover_line_dash_offset), this group of options being shown in the second position in the signature. 95 is a really high number of keywords, and is nothing compared to e.g. .heatmap() with its 217 style options for a total of 268 keyword arguments !

• Style options should be put last
• See if there's an easy way to reduce the number of these style options for some. It is likely no one ever used hover_line_dash_offset with hvPlot (it's undocumented in hvPlot and in HoloViews), should we keep it?
• Even reduced, there will still be a pretty large number of keywords (for comparison, pd.read_csv has 49). Sorting them by name isn't an option, it would be really messy. Some grouping is already in place, let's see if we can improve that.
• Geographic options should be added when relevant.