archivebox add and other entry points now seed Crawl.urls as
CrawlSeed JSONL at depth=0 (the input layer) with max_depth=depth
for direct URLs and depth+1 only for stdin/import text where the
synthetic archivebox://internal root lives at depth=0. The runner
also accepts one plain URL per line for ORM/crawl-create/schedule
callers so every Crawl row goes through the same expansion path
without scattering CrawlSeed knowledge across the codebase.
Tests updated to match restored convention.
Direct URL inputs from CLI/UI/API now seed Crawl.urls as explicit
{type:CrawlSeed,url,depth} JSONL rows; raw stdin/UI/API import text
stays verbatim. The runner's create_initial_snapshots() is now the
single place that either expands seed rows or creates the synthetic
archivebox://internal root + staticfile/stdin.txt, so add paths no
longer perform DB/FS side effects and the parser hooks run through
the same Snapshot lifecycle as every other extractor.
