remove SmallVec and LazyIndexMap from json value

[davidhewitt]

This simplifies the jiter public API (particularly on JsonValue) by simplifying the JsonObject and JsonArray types to just use Arc<Vec<...>> rather than smallvec in the array case and LazyIndexMap in the object case. This hasn't affected the way that we parse or the algorithmic complexity, it's a pure simplification.

At the same time we completely remove LazyIndexMap which attempted to be a thread-safe data structure which lazily constructed a mapping for iteration to guarantee uniqueness. I'm reasonably sure it became a "jack-of all trades, master of none" situation and there are bugs in it, maybe also performance costs.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.36%. Comparing base (4cefa8b) to head (9fabe08).
Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #184      +/-   ##
==========================================
- Coverage   88.71%   88.36%   -0.36%     
==========================================
  Files          13       12       -1     
  Lines        2207     2106     -101     
  Branches     2207     2106     -101     
==========================================
- Hits         1958     1861      -97     
+ Misses        153      151       -2     
+ Partials       96       94       -2     

Files with missing lines	Coverage Δ
crates/jiter/src/lib.rs	93.33% <ø> (ø)
crates/jiter/src/value.rs	81.02% <100.00%> (-0.09%)	⬇️

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4cefa8b...9fabe08. Read the comment docs.

[codspeed-hq]

CodSpeed Performance Report

Merging #184 will improve performances by 11.14%

_{Comparing dh/simpler-value (9fabe08) with main (11ef20f)}

Summary

⚡ 1 improvements
✅ 69 untouched benchmarks
⁉️ 3 dropped benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⁉️	lazy_map_lookup_1_10	9.2 µs	N/A	N/A
⁉️	lazy_map_lookup_2_20	25.8 µs	N/A	N/A
⁉️	lazy_map_lookup_3_50	52.4 µs	N/A	N/A
⚡	medium_response_jiter_value	41.8 µs	37.6 µs	+11.14%

[davidhewitt]

@samuelcolvin I have tested this all the way up the stack e.g. in pydantic/pydantic#11439

I am satisfied that the performance here is fine and we should merge this. In core we can do fancy iteration if we want, but they don't add much benefit and do add a lot of complexity so I am not rushing to merge that. (Eventually we might want the core changes if we move to jiter iteration in core.)

[samuelcolvin]

LGTM