fix(cli): apply the documented 600s read timeout, and report a timeout as one (#42)

[jrosskopf]

Closes #42.

Two defaults, and the lower one won

AdtSessionOptions::read_timeout defaulted to 120s while AppConfig::timeout_seconds documented 600. Both CLI entry points assign read_timeout only under if (HasFlag(args, "timeout")), so unless --timeout was passed the session kept its own 120s — for the CLI, the legacy deploy path and the MCP server alike.

Anything slower failed at 121s. The ABAP behind it keeps running server-side and normally completes, so the caller was told the work failed when it had succeeded, and had to go query the result to find out which.

There is now one constant, kDefaultReadTimeoutSeconds, behind both.

A timeout that says so

before:  Error: Post
           HTTP request failed: Failed to read connection
           Unexpected? Please report it: https://github.com/DataZooDE/erpl-adt/issues

after:   Error: Post
           request timed out after 5s (connecting to http://localhost:50000)
           Hint: Raise it with --timeout <seconds>. Work already started on the server
                 may still be running and may complete — check before repeating it.

The five places that shaped a transport failure by hand now share one MakeTransportError. Only the timeout family gets the new wording; every other transport error is byte-for-byte what it was. PrintError drops the issue-tracker line for ErrorCategory::Timeout — that invitation is for surprises, not for a limit the caller configured and can raise.

--timeout's help now states the default and that it is a per-request read timeout, in all three places that describe it, plus docs/spec2.md, which still said 120.

Verified live

Against a4h, with exactly the class from the issue (WAIT UP TO 170 SECONDS):

$ erpl-adt ... object run ZCL_ERPL_WAIT          # no --timeout
WAITED 170
rc=0 after 170s                                   # previously: failed at 121s

$ erpl-adt ... --timeout 5 object run ZCL_ERPL_WAIT
Error: Post
  request timed out after 5s (connecting to http://localhost:50000)
  Hint: Raise it with --timeout <seconds>. ...
rc=10 after 5s

1355 unit tests pass, 5 of them new: the two defaults agreeing, a real socket timing out through the LocalServer harness, a connection refusal not being dressed up as a timeout, and the issue-hint carve-out in both plain and color mode. Integration smoke suite passes.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}