perf: pre-allocate buffers with `with_capacity()`

[RoloEdits]

Pre-allocating removes a lot of regrows for the buffers.

Benchmarks

Quick benchmark opening a 150k row, 17 column .xlsx file shows 15% increase in performance.

master:

868,545,530 ns/iter (+/- 7,365,438)

perf/pre_alloc_buf:

750,808,050 ns/iter (+/- 6,366,129)

Profiles

Purple highlight is alloc.

Current master:

Change:

Changed from vec![] macro to with_capacity() as there was an introduction to a lot of instructions running.

vec![0;N]:

with_capacity(N):

[RoloEdits]

Marginal improvement but still an improvement:

pre 2881d13:

post:

[tafia]

Awesome, thanks!