refactor(tests): split oversized modules and consolidate mocks

[spalen0]

The 2,411-line explainer test module mixed prompt construction, output parsing, address handling, transaction flows, state reads, and batch execution. Split it into six focused modules under tests/ai_explainer/, and extract Telegram, alert, dispatch, and Web3 tests from test_utils.py into dedicated files.

Consolidate repeated LLM provider and HTTP response mocks, share address-test setup, and correct the missing-Telegram-credentials test to assert that no POST request is sent. Preserve every existing test case and its assertions, with that intentional correction. Shared builders have public names, and response mocks explicitly configure raise_for_status. The cleanup removes 75 lines overall.

Validation:

• Full suite: 1,691 passed, 4 skipped, 48 subtests passed.
• Review follow-up: all 143 explainer and Telegram tests passed, with 5 subtests passed.
• All 193 moved test methods and their assertions compared against the originals.
• uv run ruff check .
• uv run ruff format --check .
• uv run ty check
• git diff --check