v0.10.1 - #174

• version [prerelease]: 0.10.1a0
• update readme: generic framing and demo version
• update locations demo to save geotargets to demo dir
• plan [add]: 050 - move-parsers-to-package
• plan [activate]: 050 - move parsers.py into a parsers/ package
• move parse modules into a parsers package
• plan [log]: 050 - record pr and circular-import fix
• update parsers init to drop circular-import workaround
• drop snakeviz dev dep in favor of pstats
• drop ipykernel dev dep to fully evict tornado
• update gitignore: scope .claude dir, allow CLAUDE.md
• plan [close]: 050 - move parsers into a package
• rename parsers.parsers module to parse_serp
• harden parse: classify-first pass, to_dict copy, dedup sibling
• plan [log]: 044 - selector-pushdown negative result
• docs: add copilot agent instructions
• version [patch]: 0.10.1

Patch release. Internal restructure plus a parse-pipeline hardening pass; no public API or output change.

• Parsers package: the flat parsers.py + component_parsers/ modules are reorganized into a single WebSearcher.parsers package (parse_serp, component, component_list, component_types, components/). Public entrypoints are unchanged (import WebSearcher; WebSearcher.parse_serp, from WebSearcher.parsers.parse_serp import parse_serp); deep imports of the old flat paths must switch.
• Parse hardening: every component is now classified before any is parsed, so the two DOM-mutating parsers (general, ai_overview) can no longer make classification depend on parse order; Component.to_dict() returns a shallow copy so callers can't mutate the live attribute table; the local next-sibling walk is replaced by the shared _slx.next_sibling helper. Output is byte-identical (snapshot suite pins it).
• Dev tooling: dropped the snakeviz and ipykernel dev deps (evicting tornado); README generic framing and demo-version alignment; locations demo writes geotargets into the demo dir.