build: reduce artifact storage and release build time

[nmrtist]

Summary

• Reduce Cargo artifact storage as before.
• Set the release profile to lto = false and codegen-units = 16.
• Keep debug = 0 and strip = true.

Why

The previous thin LTO + codegen-units = 1 configuration imposed a large
incremental-build cost with only a very small end-to-end benefit on a real PlotX
2D NMR processing workload.

On Windows, using the 2D dataset (512 × 2048 complex samples),
zero-filling the direct dimension to 32k, applying 2D FFT, and exporting SVG:

Release configuration	Clean build	Small-change rebuild	Workflow mean
lto=false, codegen-units=16	270.750 s	21.763 s	1.592 s
lto=false, codegen-units=1	619.220 s	89.313 s	1.547 s
lto="thin", codegen-units=16	258.345 s	97.285 s	1.556 s
lto="thin", codegen-units=1	598.494 s	185.429 s	1.542 s

Compared with the prior shipping profile, the selected configuration reduces
clean-build time by 54.8% and small-change rebuild time by 88.3%. Its measured
end-to-end 2D workflow cost is approximately 50 ms (3.3%) per run.

Validation

• Ran the reference-backend release workspace test suite for all four profile
  combinations; all passed.
• Excluded plotx-datafusion and plotx-substrait, matching the repository's
  default CI/reference-executor scope.
• Verified the final Cargo manifest with locked Cargo metadata resolution.