Copying characters one by one is very inneficient, especially since each
time the buffer capacity has to be checked again and again.

Using `memcpy` for chunks of bytes that don't need any escaping, we can
save a lot of time.

Before:

```
ruby 3.3.4 (2024-07-09 revision be1089c8ec) +YJIT [arm64-darwin23]
Warming up --------------------------------------
            2k-plain    23.802k i/100ms
       2k-few-escape    22.693k i/100ms
Calculating -------------------------------------
            2k-plain    239.746k (± 0.5%) i/s    (4.17 μs/i) -      1.214M in   5.063403s
       2k-few-escape    227.900k (± 0.6%) i/s    (4.39 μs/i) -      1.157M in   5.078454s
```

After:

```
Warming up --------------------------------------
            2k-plain    94.975k i/100ms
       2k-few-escape    61.998k i/100ms
Calculating -------------------------------------
            2k-plain      1.039M (± 1.0%) i/s -      5.224M in   5.025882s
       2k-few-escape    658.756k (± 1.3%) i/s -      3.348M in   5.082975s
```

Bench:

```ruby

require "trilogy"
require "benchmark/ips"
t = Trilogy.new(database: "test")

plain = "A" * 2000
few_escapes = (("a" * 19) + "\n") * 100
Benchmark.ips do |x|
  x.report("2k-plain") { t.escape(plain) }
  x.report("2k-few-escape") { t.escape(few_escapes) }
end
```