lnk_points_snap(): wrap or retire in favour of the bulk fresh::frs_point_snap() #314

fresh#247 makes fresh::frs_point_snap() a bulk snap (fresh 0.40.0). It takes a table, a data frame or an sf object and runs one lateral nearest-neighbour query. With to = it writes the result to a table. It returns watershed_group_code, and its candidates and measures match lnk_points_snap(). On all 19,905 PSCIS crossings at 150 m:

• 18,180 picks are identical.
• 22 are equidistant ties: a point sits on the vertex between two segments of the same stream, and the measures land 1 m apart.
• 11 picked a placeholder segment (lnk_points_snap() can snap to placeholder (999.*) stream segments #313).
• There are no other differences. At num_features = 5 the candidate sets and measures agree too.

The comparison script is data-raw/point_snap_parity_check.R in fresh.

Proposal

Make lnk_points_snap() a thin wrapper, or deprecate it and call fresh directly:

fresh::frs_point_snap(conn, table_in, col_id = "<id>",
  tolerance = snap_tolerance, num_features = num_features,
  col_blk = blue_line_key_col, stream_order_min = stream_order_min,
  exclude_edge_types = exclude_edge_types, to = table_out)

Then join pts.* back on the id if callers need the input columns.

Differences to handle

• fresh returns blue_line_key and geom. link returns snapped_blue_line_key and geom_snapped.
• fresh needs a unique, non-missing id column (col_id) for a table input.
• A missing per-row hint leaves that point unconstrained in fresh. In link, s.blue_line_key = pts.<col> drops the row.
• fresh breaks distance ties on linear_feature_id. link has no tie-break.
• The roxygen at R/lnk_points_snap.R:52 refers to fresh::frs_point_snap_knn, which no longer exists.

Relates to NewGraphEnvironment/fresh#247, #283.