Pin dev toolchain to the versions the gates are validated against

CI resolved ruff 0.6.9 from the "^0.6" constraint while local development ran
0.15.22, so the two were enforcing different rule sets. CI failed on UP038
(prefer "X | Y" over a tuple in isinstance) -- a rule since retired upstream,
because the union form is slower at runtime than the tuple it replaces. The
code was never wrong; the linters simply disagreed.

Pin pytest, pytest-cov, mypy, ruff and hypothesis to the majors actually in
use. Caret constraints on ruff (0.x) keep a future 0.16 rule set from breaking
CI unprompted.

Also drop respx: declared as a dev dependency but imported nowhere, and not
installed in the working environment.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: SAKMZ

pyproject.toml

@@ -24,12 +24,15 @@ issues = "https://github.com/SAKMZ/n8n-python-sdk/issues"
 packages = [{ include = "n8n_iac" }]
 
 [tool.poetry.group.dev.dependencies]
-pytest = "^8.2"
-pytest-cov = "^5.0"
-mypy = "^1.11"
-ruff = "^0.6"
-hypothesis = "^6.100"
-respx = "^0.21"
+# Pinned to the majors the gates are actually validated against. Floating
+# these apart from the developer toolchain means CI lints with a different
+# rule set than local runs -- which is exactly how UP038 (since retired
+# upstream) failed CI on code that passed locally.
+pytest = "^9.1"
+pytest-cov = "^7.1"
+mypy = "^2.3"
+ruff = "^0.15"
+hypothesis = "^6.157"
 
 [build-system]
 # poetry-core 2.x is the floor: this project declares metadata in the PEP 621