perf: Optimize array_to_string to avoid a copy

[neilconway]

Which issue does this PR close?

N/A

Rationale for this change

array_to_string used a temporary buffer to convert array elements to a string, before then copying that buffer into a StringBuilder via append_value. We can skip this copy by arranging to write directly into the StringBuilder's buffer because it implements fmt::Write. This is a 12-18% performance improvement on the array_to_string` benchmarks.

What changes are included in this PR?

• Optimization
• Change the signature of the append closure to return Result, which avoids a lot of unwrap clutter

Are these changes tested?

Yes -- covered by existing tests and benchmarks.

Are there any user-facing changes?

No.

[neilconway]

Benchmarks:

  ┌──────────────────┬─────────────┐
  │    Benchmark     │ Improvement │
  ├──────────────────┼─────────────┤
  │ int64/5          │ -16%        │
  ├──────────────────┼─────────────┤
  │ int64/20         │ -13%        │
  ├──────────────────┼─────────────┤
  │ int64/100        │ -12%        │
  ├──────────────────┼─────────────┤
  │ float64/5        │ -14%        │
  ├──────────────────┼─────────────┤
  │ float64/20       │ -13%        │
  ├──────────────────┼─────────────┤
  │ float64/100      │ -12%        │
  ├──────────────────┼─────────────┤
  │ string/5         │ -18%        │
  ├──────────────────┼─────────────┤
  │ string/20        │ -15%        │
  ├──────────────────┼─────────────┤
  │ string/100       │ -14%        │
  ├──────────────────┼─────────────┤
  │ nested_int64/5   │ -13%        │
  ├──────────────────┼─────────────┤
  │ nested_int64/20  │ -13%        │
  ├──────────────────┼─────────────┤
  │ nested_int64/100 │ -12%        │
  └──────────────────┴─────────────┘

[neilconway]

cc @goldmedal , if you're interested in reviewing another small tweak to array_to_string 😊