Split adapter.py monolith into focused modules; restore Ruff complexity defaults (issue #12)

adapter.py was a 3,493-line monolith. Split it into focused modules with
clear separation of responsibilities, bringing it down to ~857 lines:

- Module-level helpers: hooks, classify, pickers, commands, setup,
  defaults, secrets
- Frame-handler mixins: inbound, tool_frames, push_frames, media_frames,
  picker_frames, channel_frames, query_frames
- mixin_base: IrisAdapterBase (declaration-only base for shared attrs)
- adapter.py now holds only IrisAdapter (the composition of the 7 mixins
  + BasePlatformAdapter), register(), and test-facing re-exports

The mixins come before BasePlatformAdapter in the MRO so their methods
override the base; super() calls (e.g. send_image) still resolve to
BasePlatformAdapter. No circular imports; dispatch.py and http_server.py
(instance-method callers) are unaffected.

Ruff complexity ceilings (PLR0911/0912/0913/0915) restored to Ruff's
built-in defaults (12/50/6/5) instead of "just above the current maxima",
which ratchets the bar down as code grows. The existing genuinely-complex
functions (frame builders mirroring the wire schema, the QR matrix builder,
the dispatch table) carry an explicit `# noqa: PLR09xx` marking them as
reviewed, frozen exceptions; new code is held to the default ceilings.

All 125 tests green (94 test_android + 31 test_android_http); no new ruff
errors introduced.

gateway-plugin/ruff.toml

@@ -4,11 +4,17 @@
 #   hermes-agent/.venv/bin/python -m ruff check gateway-plugin
 #
 # The rule set is deliberately broad (pycodestyle, pyflakes, isort, pyupgrade,
-# bugbear, flake8-simplify, pylint, return, comprehensions). Thresholds below
-# reflect the plugin's real shape: it is a single large dispatch surface
-# (adapter.py) plus a wire-protocol layer (protocol.py) whose frame builders
-# mirror the schema, so the complexity ceilings are set just above the current
-# maxima rather than an idealized small-function target.
+# bugbear, flake8-simplify, pylint, return, comprehensions).
+#
+# The pylint complexity ceilings (PLR0911/0912/0913/0915) are left at Ruff's
+# built-in defaults (see [lint.pylint]). We deliberately do NOT raise them to
+# "just above the current maxima": that ratchets the bar down every time code
+# grows (LLM maintenance adds functions, it does not refactor them), so new
+# complex code would silently pass. Instead, the handful of genuinely complex
+# functions that already exist (frame builders that mirror the wire schema,
+# the QR matrix builder, the dispatch table) carry an explicit
+# `# noqa: PLR09xx` marking them as reviewed, frozen exceptions. New code is
+# held to the default ceilings.
 
 line-length = 100
 
@@ -32,12 +38,12 @@ select = [
 ignore = ["PLC0415"]
 
 [lint.pylint]
-# Current maxima in the codebase: 22 branches, 64 statements, 9 returns,
-# 8 args (protocol.py:252 frame builder is the lone 11-arg outlier, noqa'd).
-max-branches = 24
-max-statements = 70
-max-returns = 9
-max-args = 8
+# Ruff's built-in defaults. Existing outliers are noqa'd at the def line
+# (search for `# noqa: PLR09`), not absorbed into a raised ceiling.
+max-branches = 12
+max-statements = 50
+max-returns = 6
+max-args = 5
 
 [lint.per-file-ignores]
 # The e2e / ws_probe drivers are assertion scripts: scenario numbers and