Allow `copy()` to save directly to disk via a `urlpath` parameter

[Jacc4224]

Background

copy() already performs a deep, compacted in-memory copy of a CTable. save() writes an in-memory table to disk but blocks on views. There is no single call that copies a table (or a view) directly to a new path on disk.

Proposed behaviour

• Add an optional urlpath parameter to copy().
• When urlpath is provided, the copy is written to disk instead of memory, equivalent to copy() followed by save().
• This naturally unblocks saving views to disk: a view copied with urlpath produces a new standalone compacted table containing only the view's rows and columns.

Example

import blosc2 as b2

t = b2.CTable(Row)
t.extend(rows)

# In-memory copy (current behaviour, unchanged)
t2 = t.copy()

# Copy directly to disk (proposed)
t.copy(urlpath="my_table.b2d")

# Also works on views — saves only the filtered rows/columns
view = t.where(t["price"] > 100)
view.copy(urlpath="filtered_table.b2d")

Notes

This makes saving views to disk a first-class operation without needing to relax the block in save().