Add to= parameter to frs_network() for DB-side pipeline

[NewGraphEnvironment]

Problem

frs_network() returns sf objects to R memory. All downstream pipeline functions (frs_col_join, frs_col_generate, frs_break, frs_classify, frs_aggregate) operate on DB tables. This forces a round-trip through R between the network query and the pipeline — at province scale (multiple watershed groups), materializing all geometries in R blows up memory.

Proposed Solution

Add to parameter. When provided, write results to DB tables instead of returning sf:

# Current: pulls to R (unchanged, to=NULL default)
result <- frs_network(conn, blk, drm, tables = list(streams = "..."))

# New: stays on DB, returns conn for piping
conn |>
  frs_network(blk, drm, to = "working.byman", tables = list(
    streams = "whse_basemapping.fwa_stream_networks_sp",
    lakes = "whse_basemapping.fwa_lakes_poly")) |>
  frs_col_join("working.byman_streams", ...) |>
  frs_col_generate("working.byman_streams") |>
  frs_break("working.byman_streams", ...) |>
  frs_classify("working.byman_streams", ...)

Multi-table naming

to is a prefix. Table list names become suffixes:

• to = "working.byman" + streams → working.byman_streams
• to = "working.byman" + lakes → working.byman_lakes
• Single table: to used as-is (no suffix)

Internal changes

• frs_network_direct() and frs_network_waterbody() get to param
• When to is NULL: frs_db_query() → returns sf (current)
• When to is provided: .frs_db_execute("CREATE TABLE AS") → returns NULL
• SQL body identical — only execution path changes
• overwrite param (default TRUE) for DROP IF EXISTS
• Error if clip and to both provided (clip is R-side spatial op)

Blast radius

All existing callers use to=NULL (default) — zero breakage:

• fresh tests (35+), vignettes (4), breaks app_server.R, restoration scripts (3)

Only data-raw/example_byman_ailport.R and the new habitat pipeline vignette will use to=.

[NewGraphEnvironment]

to= parameter added in PR #56. Commit aac96ef.