Feat/cli helpers header parity

[pravirkr]

Summary

Port the remaining SIGPROC IO / reduction / fake tools into thin spp_* CLIs on top of the existing sigpyproc API. We aim to make sigpyproc a modern successor, not a C/Fortran clone; functional equivalence is prioritized over bit-identical numerics (except for byte-copy I/O). A modern search API and spp_search CLI (FFT + harmonics, FFA, single pulses, harmonic clustering).

Should close #18.

What's included

Inspect and extract

• spp_header: print (sigproc / pfits / fbh5), get, keys-help, in-place update, and --dry-run Before/After table
• spp_extract: time range (--start-sec / --duration), channel list or range; extract_chans now sets per-channel fch1
• spp_stats: per-channel mean / std / skew / kurtosis (text or CSV)

Clean and reduce

• Filterbank.requantize actually packs bits (Rescale + BitsInfo.quantize)
• spp_decimate, spp_bandpass, spp_zerodm (Eatough / PRESTO weighted zero-DM)
• spp_clean: RFI mask, zap channels, zap times (zap_block in core/rfi.py)

Combine

• sigpyproc.combine: time-merge and frequency-splice (lives outside io/ so readers are not imported there)
• spp_merge, spp_splice
• FREQUENCY_START headers are detected and rejected with a clear error (linear fch1/foff only)

Dedisperse and fold

• spp_dedisperse tim | subband | tree | plan
• Chunked Taylor tree (kernels.taylor_tree) and Levin DM plan (core/dmplan.py)
• spp_fold writes .npz + ASCII profile

Fake data

• simulation.pulsar + spp_fake {noise,pulsar,pulse}
• N(0,1) + rectangular pulse, quantized without a running rescale so injected SNR is kept
• Furby stays the FRB injector

Search

• CandidateTable (QTable, SI units, CSV only)
• PowerSpectrum.find_peaks — running median/MAD SNR, local maxima
• TimeSeries.search - resample → rfft → FourierSeries.deredden → form_spec → harmonic folds → find_peaks
• TimeSeries.find_pulses - boxcar bank + greedy dead-time (not oneMatchedFilter call, not peak.f time-halving)
• TimeSeries.ffa + kernels.ffa_stage — Staelin/Mueller tree; periods in seconds on the API
• CLI: spp_search {fft,pulses,ffa,cluster}

Tutorial

• docs/tutorials/search.md, wired into Get Started

SIGPROC mapping

SIGPROC	This PR
seek / find	TimeSeries.search / spp_search fft
peak	TimeSeries.find_pulses / spp_search pulses
best (non-plot)	CandidateTable.cluster / spp_search cluster
ffa	TimeSeries.ffa / spp_search ffa
.prd / .bst / PGPLOT	not implemented (CSV instead)

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB