Provide an API for performant bulk or sequential read/writes

[fzhinkin]

As stated in #132, kotlinx-io don't include Okio APIs aimed for performant sequential and bulk ops (like, UnsafeCursor).

Such API is required to implement previously removed functionality (like select) efficiently. It's also required for integration with platform-specific APIs (like hash functions implementation using Java's MessageDigest API).

This issue claims the intent to provide such functionality, but the actual API will be discussed later.

[whyoleg]

Just wanted to mention, that from my perspective this should be implemented taking into account future feature from #131: different memory types (in addition to byte arrays) Buffers are built upon (like NIO ByteBuffers on JVM);
Use-case: ktorio/ktor-io#7
Ideally it should be possible to use JDK MessageDigest/Cipher API for Buffer with multiple underlying segments of different types like ByteBuffer, ByteArray, MemorySegment and even Netty ByteBuf (which allows to get ByteBuffer from it).

[fzhinkin]

An implementation will definitely take different memory types into account, otherwise, it'll be hard to make performant reads/writes.

[rnett]

I am trying to implement a custom HashingSink (using Guava's hash functions, but it doesn't really matter) and found that implementing a custom sink is rather difficult because it's not easy to read (groups of) bytes from a Buffer yourself, especially without "consuming" them (so I can forward to the "real" sink). This makes it very hard to implement custom sinks or sources. It seems like this issue would be the feature responsible for solving this, is that correct? If not I'll create a new one.

[fzhinkin]

@rnett, that's correct. Once solved, this issue should allow more easily implementing things like HashingSink. I'm currently finalizing the design and will post it soon.