Clean up and modernize code for dedupe

[jonkoops]

Cleans up the code for dedupe so that it better matches the structure of the other variants. Also introduces Set to de-duplicate classes. Benchmarks seem to indicate it's slightly faster, especially for mixed test cases:

Node.js (v21.5.0)

Benchmarking 'strings'.

Task Name	ops/sec	Average Time (ns)	Margin	Samples
dedupe/local	4,045,085	247.2135826039219	±0.56%	2022543
dedupe/npm	4,042,492	247.37213314357965	±0.52%	2021247

Benchmarking 'object'.

Task Name	ops/sec	Average Time (ns)	Margin	Samples
dedupe/local	7,221,118	138.48269326646596	±0.59%	3610560
dedupe/npm	6,935,808	144.1793010477904	±0.29%	3467905

Benchmarking 'strings, object'.

Task Name	ops/sec	Average Time (ns)	Margin	Samples
dedupe/local	4,415,599	226.46980115941534	±0.21%	2207800
dedupe/npm	4,096,954	244.08371678890256	±0.74%	2048478

Benchmarking 'mix'.

Task Name	ops/sec	Average Time (ns)	Margin	Samples
dedupe/local	3,083,596	324.2966677236706	±0.74%	1541799
dedupe/npm	1,935,753	516.594635475142	±0.76%	967877

Benchmarking 'arrays'.

Task Name	ops/sec	Average Time (ns)	Margin	Samples
dedupe/local	1,921,840	520.3345623628504	±0.39%	960921
dedupe/npm	1,888,877	529.4150696866383	±0.95%	944439

[dcousens]

Concept LGTM, I'll need to review later, happy if you want to merge and I'll come back to this 💛

[jonkoops]

I need to give this a little more testing, I am seeing some regressions in Firefox and Chrome, not entirely sure why so I'll have to do some profiling.

[jonkoops]

Closing this because I am seeing some inconsistencies in performance. I'll refactor the existing code in a structure that is easier to work on and then refactor it piece by piece.