fix(telegram): pass proxy URL explicitly to HTTPXRequest when proxy env vars are set

[MaybeRichard]

Problem

When HTTPS_PROXY / HTTP_PROXY / ALL_PROXY env vars are set, the gateway correctly detects proxy_configured=True and skips the fallback-IP transport — but then creates HTTPXRequest(**request_kwargs) without an explicit proxy= argument, relying solely on httpx's trust_env mechanism.

In practice, trust_env alone is unreliable when the proxy is an HTTP CONNECT proxy (e.g. Clash / ClashMac in fake-ip mode, common for users in China): the TLS handshake inside the CONNECT tunnel fails intermittently via httpcore's anyio backend, even though the same proxy works fine from curl or requests/urllib3. The result is a flood of httpx.ConnectError / httpx.ConnectTimeout errors and failed message delivery.

Fix

When proxy_configured is True, read the proxy URL from the environment and pass it explicitly via HTTPXRequest(..., proxy=proxy_url). This matches the explicit approach already used in the aiohttp-based adapters (weixin/wecom/matrix, see #8680) and is more robust than trust_env for users behind HTTP CONNECT proxies.

# before
request = HTTPXRequest(**request_kwargs)
get_updates_request = HTTPXRequest(**request_kwargs)

# after
proxy_url = (
    os.getenv("HTTPS_PROXY") or os.getenv("https_proxy") or
    os.getenv("HTTP_PROXY") or os.getenv("http_proxy") or
    os.getenv("ALL_PROXY") or os.getenv("all_proxy")
)
request = HTTPXRequest(**request_kwargs, proxy=proxy_url)
get_updates_request = HTTPXRequest(**request_kwargs, proxy=proxy_url)

Test

Verified on macOS with ClashMac proxy (http://127.0.0.1:7890) in fake-ip mode. Before the fix: repeated ConnectError / TimedOut on polling and sends. After the fix: stable connection, messages delivered reliably.

🤖 Generated with Claude Code

[MaybeRichard]

The failing tests (test_telegram_photo_interrupts, test_session_hygiene, etc.) are pre-existing failures on origin/main — confirmed by running the same test suite against the upstream code before applying this PR's change. They are unrelated to the proxy fix.

The build-and-push failure is also a known flaky CI issue (npm install network timeout, exit code 254) that appears on main branch runs as well.

[teknium1]

Merged via PR #8981. Your commit was cherry-picked onto current main with your authorship preserved in git log. Used the shared resolve_proxy_url() from base.py instead of inline env var reading (picks up macOS system proxy for free too). Thanks for the fix!