Per-test `rpad` and per-file clock reads still go through `$( )` instead of a return slot

[Chemaclass]

Summary

Two spots still capture a return value through $( ) where the house return-slot pattern already exists next door.

Current behavior

Per passing test. src/console/test_line.sh:33-37:

full_line="$(bashunit::str::rpad "$line" "$_BASHUNIT_CONSOLE_DURATION_OUT")"

bashunit::str::rpad (src/util/str.sh:92) is already fork-free inside. The subshell around it is the whole cost. Its neighbours already have slot variants: strip_ansi_to_slot (src/util/str.sh:23) and format_duration_to_slot (src/console/duration.sh:11).

This one fires on Bash 5, not on 3.2, because auto disables per-test timing when the clock is a perl fork (src/config/env.sh:524-529, src/util/clock.sh:95).

The failure renderers have the same shape at src/console/test_line.sh:71, :123, :192, :207, :223 and :256, on the failure path only.

Per file with a file hook. src/runner/hooks.sh:172, :189, :423, :440:

local start_time
start_time=$(bashunit::clock::now)

bashunit::clock::now_to_slot (src/util/clock.sh:111) is the fork-free variant, and run_test was already converted to it (src/runner/exec.sh:542, :599). These four were missed. On Bash 3.2 each one is a subshell wrapping a perl fork, twice per set_up_before_script and twice per tear_down_after_script. This suite has 115 files with the former.

bashunit::clock::init (src/util/clock.sh:234) captures $(bashunit::clock::now) the same way, once per run.

Expected behavior

Add bashunit::str::rpad_to_slot and use it. Switch the four hook clock reads and clock::init to now_to_slot.

Small next to #1 and #2, and grouped here because it is the same one-line change four times over. Unit-test rpad_to_slot against rpad on the same inputs, including an ANSI-coloured string and one over 1024 chars, since that is where strip_ansi_to_slot changes strategy (src/util/str.sh:82).

[Chemaclass]

Done in #1358 (merged).