ARROW-9777: [Rust] [IPC] write custom metadata [WIP]

[mqy]

I think this PR is a sub task of issue ARROW-9777.
Sorry had not seen this issue before I finished this PR.

This PR is extracted from one of my dev branch, which has more changes for custom metadata. You may have a look at the diff to arrow master.

Two design choices:

Added a new type CustomMetaData (alias to BTreeMap), in datatypes.rs.

Why BTreeMap instead of HashMap? Because Field implements traits Hash, PartialOrd, and Ord,
but HashMap breaks the behavior.

Additional refactoring

It tends to be tedious when add more args, for example:
try_new, try_new_with_options, try_new_with_options_and_custom_metadata.

Finally, I refactored several functions according to builder pattern. Sorry that, this refactoring may introduce some inconveniences.

[github-actions]

https://issues.apache.org/jira/browse/ARROW-9777

[codecov-io]

Codecov Report

Merging #9011 (f13c00b) into master (a4f7c4a) will decrease coverage by 0.04%.
The diff coverage is 51.06%.

@@            Coverage Diff             @@
##           master    #9011      +/-   ##
==========================================
- Coverage   82.87%   82.82%   -0.05%     
==========================================
  Files         201      201              
  Lines       49739    49781      +42     
==========================================
+ Hits        41220    41232      +12     
- Misses       8519     8549      +30     

Impacted Files	Coverage Δ
rust/arrow-flight/src/utils.rs	0.00% <0.00%> (ø)
rust/arrow/src/datatypes.rs	76.45% <ø> (ø)
...ntegration-testing/src/bin/arrow-file-to-stream.rs	0.00% <0.00%> (ø)
...ion-testing/src/bin/arrow-json-integration-test.rs	0.00% <0.00%> (ø)
...ntegration-testing/src/bin/arrow-stream-to-file.rs	0.00% <0.00%> (ø)
rust/arrow/src/ipc/writer.rs	76.88% <55.69%> (-6.37%)	⬇️
rust/arrow/src/ipc/reader.rs	82.90% <100.00%> (+0.03%)	⬆️
rust/parquet/src/arrow/schema.rs	90.62% <100.00%> (ø)
rust/arrow/src/array/transform/fixed_binary.rs	78.94% <0.00%> (-5.27%)	⬇️
rust/arrow/src/ipc/gen/Message.rs	34.37% <0.00%> (+2.34%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a4f7c4a...f13c00b. Read the comment docs.

[mqy]

I just mark this PR as draft, because I think there must have some missing tests to add.

@nevi-me since this PR may overlaps with your existing work, please let me know should I close this PR or not?

[nevi-me]

Hey @mqy, I'm on vacation, so I haven't been checking the project much.

ARROW-10299 is for implementing ipc::MetadataVersion::V5 on the write side (reads are mostly fine)
ARROW-10258 depends on ARROW-10259 that you're working on.

[nevi-me]

The major downside with a new() that doesn't create the header, is that it creates a burden on end-users to remember to call the write_header_schema(). I would prefer to stick with try_new() and try_new_with_options()(happy to rename the latter), where we now amend the latter to also take other parameters.

I don't mind us using a builder pattern, but we shouldn't change the try_new in the process.

[mqy]

Agree, so add metadata to IpcWriteOptions?

[mqy]

Hey @mqy, I'm on vacation, so I haven't been checking the project much.

ARROW-10299 is for implementing ipc::MetadataVersion::V5 on the write side (reads are mostly fine)
ARROW-10258 depends on ARROW-10259 that you're working on.

@nevi-me I had seen that your status is busy.

The major reason that I closed this PR is because I don't want it in the pull request list even if in draft mode too long.
Now since I got you feedback, I'll continue this PR later after #9025