cytorete v0.1.1 — two lookups that failed on every cytome

A patch release. Nothing about the method changed; two functions that could
not work on a .cytome file now do.

Fixed

• cytorete.pl.regulonEmbedding raised Embedding not found: X_umap on
  every cytome. Conversion stores obsm arrays under a modality prefix —
  RNA_umap since cytome 0.2.6, RNA_obsm_X_umap before it — and the lookup
  was exact, so the function failed regardless of how the file was written.
  It now resolves the short name the way plotting's basis= does, matching
  both naming generations; an unknown key lists what the file actually has.

• cytorete.pl.regulonNetwork drew unlabelled targets. The TF was named
  and every target was an anonymous dot, which conveys only that the TF has
  targets — something the target count already said. Target names are drawn
  by default; pass label_targets=False for dense multi-TF layouts.

Added

• inferRegulon accepts a .cytome path as well as an open dataset,
  matching the rest of the stack, and closes what it opened. Previously a
  path raised AttributeError: 'str' object has no attribute 'var_names'
  three frames down, naming nothing the caller had done wrong.

Upgrading

pip install -U cytorete

No API changes. Code written against 0.1.0 keeps working; the two functions
above start working on inputs where they previously raised.

Suite: 60 passed in the published distribution, against released
piaso-tools / cosg / cytome.

cytorete v0.1.0 — RNA regulons: promoter cistrome × trans co-specificity

cytorete — Greek cyto- (κύτος, "cell") + Latin rēte ("net"), the
cell's network. Pronounced sy-toh-REE-tee (/ˌsaɪtoʊˈriːtiː/), with
rete as in rete mirabile. The name is the claim: a regulatory network
is a property of a cell type, not of a tissue.

cytorete infers regulons (a TF and its target set, with weights) from
scRNA-seq, on the PIASO stack:

1. Promoter cistrome — strand-aware promoter windows from the PIASO-data
   TSS BEDs, sequences from the 2bit genome, scanned with piaso.pp.scan_motifs
   (JASPAR / CIS-BP) → a TF × gene motif-support matrix.
2. inferRegulon — motif-supported edges scored by trans
   co-specificity (piaso.tl.cospecificity_trans): a TF and a target
   qualify when their COSG cell-type specificity profiles agree, per cell
   type, with NES pruning.
3. regulonActivity / regulonSpecificity — per-cell regulon scores
   (PIASO's scoring engine; AnnData or cytome streaming) and cell-type
   specificity of each regulon.
4. Plots — regulon heatmaps and TF-vs-activity scatters.

Works on AnnData or on .cytome files, in memory or streamed. inferRegulon
also takes a .cytome path directly and closes the file when it is done.
Names for the multiome (RNA+ATAC) chain exist but raise an informative
ImportError in this distribution.

Scale. One Stereo-seq section — 121,767 spatial bins × 28,204 genes —
takes 29 s for the regulon step at roughly a gigabyte of RSS, because the
matrix is streamed and never held whole.

Requires piaso-tools >= 1.2.2, cosg >= 1.1.2, cytome >= 0.2.5.

Layout. cytorete.tl and cytorete.pp hold their modules directly —
there is no grn subpackage, because the whole package is the GRN method
and the extra segment would name nothing. So it is cytorete.tl.inferRegulon,
not cytorete.tl.grn.inferRegulon.

Suite (this distribution): 60 passed, 0 skipped, against the released
piaso-tools / cosg / cytome wheels. Tests skip locally without the
optional motif extra (py2bit) or a JASPAR MEME file; the release
workflow requires the extras and fetches JASPAR, so a release run has the
full suite rather than a quietly reduced one.

Documentation lives with the rest of the stack on
piaso.org: RNA regulon inference and regulons on spatial data,
which runs the whole workflow across a mouse embryo section and checks
the result against the published SCENIC regulons for the same tissue.