feat: introduce typeCast for execute method

[wellwelwel]

This PR introduces typeCast for the execute method and allows the use both from connection level and query level.

How it works?

• Usage is strictly and precisely as it is used in the query method.
• The modifications to the binary parser were adapted from the text parser.

---

Motivation

Using the benefits of type customization without losing the advantages of the execute method.

• typeCast not supported with .execute #649 (🗓️ 2017)
• Queries run with execute don't use the global typeCast function #717 (🗓️ 2018)

About #1446 (RFC: deprecate typeCast and introduce resultsMapper), I believe that as long as typeCast is still used, it's really useful that it works for both methods.

---

How was it tested?

I duplicated and refactored all the derived TypeCast tests for query and shaped them to work in execute, as well as creating new ones.

• Creating a new test for execute that is exactly the same for each existing test on query:
  • test-type-cast-null-fields-execute.js
  • test-type-casting-execute.js
  • test-typecast-geometry-execute.js
  • test-typecast-execute.js
• Creating a new test for both query and execute, testing the typeCast on connection level and then overwriting it on query level:
  • test-typecast-overwriting.js
  • test-typecast-overwriting-execute.js
• Including a simple execute test for Bun at:
  • test-select-1.js
  • test-select-ssl.js

For benchmark comparison, see #2398 (comment).

---

Notes

Based on #649 (comment), I'll consider this a feature rather than a fix.
It's expected to close both #649 and #717.

---

It would be great if someone could check it out and please, feel free to ask and discuss anything about it 🤝

[wellwelwel]

Benchmarks

Comparing this PR and two of the website's most recent "bump" PRs (which don't interfere in the main code) using the main code's latest state.

.github/workflows/benchmark.yml

• feat: introduce typeCast for execute method
• build(deps-dev): bump @types/node from 20.11.5 to 20.11.6 in /website
• build(deps): bump lucide-react from 0.314.0 to 0.315.0 in /website

---

read 43 column definitions

(select * from mysql.user)

This PR	bump @types/node (website)	bump lucide-react (website)
{ value: 49345, range: '±0.59%', unit: 'ops/sec', extra: '98 samples' }	{ value: 46807, range: '±1.56%', unit: 'ops/sec', extra: '93 samples' }	{ value: 47599, range: '±1.06%', unit: 'ops/sec', extra: '90 samples' }

---

(select * from mysql.user) and access column lazy fields once

This PR	bump @types/node (website)	bump lucide-react (website)
{ value: 5976, range: '±0.19%', unit: 'ops/sec', extra: '99 samples' }	{ value: 4609, range: '±0.08%', unit: 'ops/sec', extra: '99 samples' }	{ value: 6041, range: '±0.14%', unit: 'ops/sec', extra: '97 samples' }

---

(select * from mysql.user) and access column lazy fields 5 times

This PR	bump @types/node (website)	bump lucide-react (website)
{ value: 4660, range: '±0.65%', unit: 'ops/sec', extra: '95 samples' }	{ value: 4537, range: '±0.32%', unit: 'ops/sec', extra: '97 samples' }	{ value: 4629, range: '±0.35%', unit: 'ops/sec', extra: '96 samples' }

[wellwelwel]

Hi, @sidorares 🙋🏻‍♂️

For this PR, I'd like it to be your decision whether or not to merge.

Although I'm optimistic about all the tests conducted, I've asked for opinions on some related Issues in an effort to get more insights/feedbacks.

[sidorares]

I think it's good to merge. Going forward, I'd like to see something better than typeCast ( an option that would allow mapping of the received data rather then controlling parsing state/logic ), but right now there is already existing code that would benefit typeCast on .execute ( mysqljs/mysql users taht would like to migrate to mysql2 + repared statements but need to use typeCast )

[sidorares]

Feel free to merge - thanks for your work @wellwelwel !!!