`Zstd.decompress` only decodes the first frame of concatenated frames

[Watson1978]

When given input containing multiple concatenated zstd frames, Zstd.decompress returns only the decompressed output of the first frame and silently drops the rest.
The zstd format explicitly supports frame concatenation (e.g. cat a.zst b.zst | zstd -d yields the concatenation of both), and the underlying ZSTD_decompress / ZSTD_decompressStream decode all frames.

Environment

• OS: Linux 7.1.1-2
• Ruby: ruby 4.0.5
• zstd-ruby: 2.0.6

Reproduction

require 'zstd-ruby'

a = Zstd.compress("Hello, ")
b = Zstd.compress("World!")
concatenated = a + b

p Zstd.decompress(concatenated)
# => "Hello, World!"      (expected)
# => "Hello, "            (actual)

Three or more frames behave the same way — everything after the first frame is lost.

Affected code

ext/zstdruby/zstdruby.c

decode_one_frame stops as soon as one frame finishes (ZSTD_decompressStream
returns 0):

if (ret == 0) {
  break;   // end of ONE frame
}

and rb_decompress returns immediately after decoding the first data frame, instead of looping back to process any remaining frames:

if (magic == ZSTD_MAGIC) {
  ZSTD_DCtx *dctx = ZSTD_createDCtx();
  VALUE out = decode_one_frame(dctx, in + off, in_size - off, kwargs);
  ZSTD_freeDCtx(dctx);
  RB_GC_GUARD(input_value);
  return out;   // <- returns; trailing frames are never read
}

The skippable-frame branch already advances off and continues the loop; the data-frame branch does not.

Suggested fix (direction)

Keep scanning until the whole input is consumed, accumulating output across frames:

• Have decode_one_frame report how many input bytes it consumed (the final
  ZSTD_inBuffer.pos) so the caller can advance off.
• In rb_decompress, append each frame's output to a single result buffer and
  advance off by the consumed length instead of returning after the first frame.