feat: Chunkify single parts to generate them in parallel

[clflushopt]

• Closes [FEATURE] Mutli-threaded generation of specific parts (when --part and --parts are specified) #80

[alamb]

Looks good to me

I ran with

cargo run --release -- --tables lineitem --scale-factor 1000 --part 1 --parts 100 --format=parquet

And it kept all my cores busy as promised.

Thank yoU @clflushopt

[alamb]

I made a PR to add some integration tests here

• feat: Add integration tests for tpchgen-cli #156

I think #156 covers the correctness aspect of this PR

I am not sure it covers the "make it parallel" functional aspect, but then again I don 't really have a great story about that

[alamb]

I wrote up some high level thoughts here

• [DISCUSS] v2.0.0 #154 (comment)

[clflushopt]

Closing this PR since the approach used doesn't guarantee consistent output.

[alamb]

I merged up from main -- there are a few test items i need to fix

[alamb]

@clflushopt -- if you have a moment perhaps you could look at this PR

I found:

1. The "split the job into multiple parts" works well (we just have to avoid Region and Nation tables) -- it passes tests well
2. The interplay of parts / part and multiple files is a little confusing
3. There is no good way to return errors

My suggested path forward is:

1. Go with the approach in this PR
2. Make a follow on refactoring PR that changes parts and part to Option<i32> and defaults them to None to make the logic clear

What are your thoughts

[clflushopt]

Sounds good @alamb did a sanity check at scale factor 10 with master it all looks good ! About your comment yes, I do agree that parts, part is a bit confusing and switching to Option<i32> could make the logic more clear.

[clflushopt]

Can be merged once you approve @alamb thanks for picking this one up !

[alamb]

Thanks @clflushopt