chore(core): growable DirectByteSink in native code via byte_sink.h

[amunra]

Technical enhancements to enable populating a byte sink from native code.

• Extracted the core of DirectByteCharSink into io.questdb.std.bytes.DirectByteSink.
• Reworked the DirectByteSink implementation by adding a layer of indirection and exposing its core as a C struct called questdb_byte_sink (see byte_sink.h).
• This is to allow passing byte sinks to native code via pointer to allow the native code to append to them without calling back into Java via JNI when it is necessary to grow the buffer.

Additionally, this PR comes with:

• Additional test cases.
• A new minimum capacity size of 32 bytes (which is the de-facto malloc minimum).
• On realloc, a more consistent new capacity calculation that finds the next power of two (see https://stackoverflow.com/questions/64734694/why-do-dynamic-arrays-have-to-geometrically-increase-their-capacity-to-gain-o1 on why this is valuable).
• Introduction of a new io.questdb.std.bytes package and inside it:
  • interface ByteSequence
    • byte byteAt(int index), int size();
    • Note that a future PR from @puzpuzpuz will remove io.questdb.std.str.ByteSequence, so this name collision is temporary.
  • interface DirectByteSequence extends ByteSequence
    • long lo(), long hi(), long ptr()
  • interface NativeByteSink extends QuietCloseable
    • provides ptr() returning the aforementioned questdb_byte_sink*
    • and a close() method to ensure correct memory bookeeping after use.
  • interface BorrowableAsNativeByteSink
    • NativeByteSink borrowDirectByteSink();
    • For types that want to expose a NativeByteSink
  • class DirectByteSink (as aforementioned), containing the core logic for DirectByteCharSink, void of any CharSink logic.
• Layered the existing {Direct}Utf8Sequence interfaces on top of the new {Direct}ByteSequence.

N.B.: DirectByteCharSink is also going to be removed and replaced by another utf-8 sink type in a future PR by @puzpuzpuz.

N.B.: ByteSequence, for now - as a compromise, has an int-based length rather than long. This is to allow for easier migration to these new interfaces.

[puzpuzpuz]

Not all binaries are updated.

[bluestreak01]

this does not look like a user facing "feature", PR title should reflect that (chore)

[amunra]

While I don't have a use for it yet, we might want to be able to transfer ownership of the sink to/from Java and native code by leaking/acquiring the impl pointer.

E.g.:

Transfering ownership from Java to native code.

DirectByteSink dbs = ..;
long ptr = dbs.leak();  // pointer can be handed over to native code.

Transfering ownership from native code to Java.

// some JNI call that constructed the buffer earlier, maybe another thread.
long ptr = ...
DirectByteSink dbs = new DirectByteSink();
dbs.of(ptr);

Something to revisit if a need arises :-)

[ideoma]

[PR Coverage check]

😍 pass : 113 / 116 (97.41%)

file detail

	path	covered line	new line	coverage
🔵	io/questdb/std/bytes/DirectByteSink.java	78	81	96.30%
🔵	io/questdb/std/bytes/DirectByteSequence.java	3	3	100.00%
🔵	io/questdb/std/str/DirectUtf8Sequence.java	3	3	100.00%
🔵	io/questdb/std/Unsafe.java	6	6	100.00%
🔵	io/questdb/ServerMain.java	4	4	100.00%
🔵	io/questdb/std/str/DirectByteCharSink.java	18	18	100.00%
🔵	io/questdb/std/MemoryTag.java	1	1	100.00%