Zero alloc indent write

[cmrschwarz]

The current with_indent function allocates a string to render the content into for every single indent_aware_write.
It also operates on a per char basis instead of processing full line slices at a time.
This PR fixes this by implementing an improved version that operates directly on the output stream.

In my testing this gives us about a 5 to 15% performance boost 🚀 ,
although the benchmarks are a bit flaky. I also added another benchmark with some
more nested partials than in the others, which benefits the most from this change.

    Running benches/bench.rs (target/release/deps/bench-a2a79c58e18ef2a8)
parse_template          time:   [20.801 µs 20.830 µs 20.858 µs]
                        change: [-0.0628% +0.2014% +0.4363%] (p = 0.12 > 0.05)
                        No change in performance detected.

render_template         time:   [3.0293 µs 3.0355 µs 3.0433 µs]
                        change: [-9.4375% -8.9483% -8.4201%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 15 outliers among 100 measurements (15.00%)
  1 (1.00%) low mild
  4 (4.00%) high mild
  10 (10.00%) high severe

large_loop_helper       time:   [225.11 µs 225.64 µs 226.28 µs]
                        change: [-5.9605% -5.6607% -5.3786%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 11 outliers among 100 measurements (11.00%)
  7 (7.00%) high mild
  4 (4.00%) high severe

large_loop_helper_with_context_creation
                        time:   [449.95 µs 450.37 µs 450.87 µs]
                        change: [-4.4517% -4.1931% -3.9515%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 13 outliers among 100 measurements (13.00%)
  12 (12.00%) high mild
  1 (1.00%) high severe

large_nested_loop       time:   [250.05 µs 250.35 µs 250.71 µs]
                        change: [-5.7467% -5.5139% -5.2769%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 8 outliers among 100 measurements (8.00%)
  6 (6.00%) high mild
  2 (2.00%) high severe

deeply_nested_partial   time:   [235.92 µs 236.23 µs 236.59 µs]
                        change: [-17.574% -17.235% -16.747%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  5 (5.00%) high mild
  4 (4.00%) high severe

[coveralls]

coverage: 81.234% (-0.3%) from 81.548%
when pulling cace44d on cmrschwarz:zero_alloc_indent_write
into 080b5e0 on sunng87:master.