High memory allocation in user-defined functions/aggregates arguments

Hi there, first of all thank you so much for this project! It's been a treat to use since switching from the mattn driver.

I help develop a tool that emulates the Google BigQuery SQL Engine by transforming queries into SQLite-compatible queries + functions. As a result, some functions get called many thousands of times per query. When using the Go memory profiler, I see a high number of allocations come from this slice initialization.

Would it be possible to mitigate this by using a sync.Pool for the slice? Do you have any other recommendations on limiting allocations at the driver-level? I could be way off here, but appreciate any thoughts you have 😄

Here's an example profile:

Showing nodes accounting for 3055.28MB, 99.15% of 3081.36MB total
Dropped 339 nodes (cum <= 15.41MB)
      flat  flat%   sum%        cum   cum%
  635.62MB 20.63% 20.63%   637.43MB 20.69%  github.com/Recidiviz/go-zetasqlite/internal.DecodeValue
  474.79MB 15.41% 36.04%  1116.15MB 36.22%  github.com/Recidiviz/go-zetasqlite/internal.setupNormalFuncMap.func1
  473.39MB 15.36% 51.40%   473.39MB 15.36%  modernc.org/sqlite.errorResultFunction
  473.01MB 15.35% 66.75%   473.75MB 15.37%  modernc.org/sqlite.funcTrampoline.func1
  417.19MB 13.54% 80.29%   684.66MB 22.22%  modernc.org/sqlite.functionArgs
  315.55MB 10.24% 90.53%  3063.42MB 99.42%  modernc.org/sqlite.funcTrampoline
  265.73MB  8.62% 99.15%   265.73MB  8.62%  modernc.org/libc.GoString