feat: Initial ORC file format support - WIP

[Jefffrey]

Which issue does this PR close?

Relates to #4980

Rationale for this change

Initial support for reading ORC files

What changes are included in this PR?

New synchronous reader of ORC files which implements RecordBatch iterator

Supported datatypes:

• Smallint
• Int
• Bigint
• Float
• Double
• String
• Char
• Varchar
• Boolean
• Tinyint
• Binary
• Date

Unsupported:

• Decimal
• Timestamps
• Compound types (nested struct, list, map, union)

Currently does simple read of ORC files into RecordBatches, no filtering support. But does support top level projection.

Supports all compression: None, zlib, snappy, lzo, lz4, zstd (delegates to other rust crates for actual decompression)

Are there any user-facing changes?

New crate

[Jefffrey]

I've marked it as WIP, but I mainly intend to clean up the code a bit, do more tests and improve docs for now

[tustvold]

Could we perhaps split this into smaller pieces of less than 1000 lines a go?

[alamb]

Could we perhaps split this into smaller pieces of less than 1000 lines a go?

Maybe we could start with the basic scaffolding for the arrow-orc crate (and CI checks, etc) and move on from there

[Jefffrey]

I've tried splitting it into a more initial version. To note that the proto stuff (file and generated rust) does make up about 1.2k of the 2.4k total lines changed, but can cut down more if desirable

@alamb @tustvold

[alamb]

Hi @Jefffrey -- thanks for the ping. I don't think I am going to have time to review this PR for a while unfortunately

[alamb]

I left a comment about review capacity here: #4980 (comment)

[Jefffrey]

Following discussion in #4980, will focus on implementing ORC file format support in https://github.com/datafusion-contrib/datafusion-orc with aim of eventually bringing into arrow-rs in the future once it's more feature complete. Closing this PR