v0.30.6 - bugs and fallbacks

• gget blat: Improved resilience against UCSC BLAT endpoint failures (fixes intermittently failing tests).
  • Added retry-with-exponential-backoff for transient failures (HTTP 429/5xx, network errors, and non-JSON 200 responses caused by UCSC rate-limiting or HTML error pages). Up to 4 attempts with 1.5s → 3s → 6s backoff.
  • Replaced the misleading "sequence too short or assembly invalid" message with the actual server response (status code, response preview) so failures are diagnosable.
  • HTTPError and URLError are now caught explicitly instead of bubbling up as unhandled exceptions.
• Bug fixes:
  • gget cosmic: Fixed misleading error message when the download step fails — was reporting the previous command's return code/stderr instead of the failing command's.
  • gget cosmic: Narrowed the JSON parse exception handler to json.JSONDecodeError so unrelated ValueErrors are no longer masked by the "Failed to download file" message.
  • gget --version, gget --help, gget invoked with no arguments, and gget <module> with no further arguments now all exit with status 0 instead of 1, so CI scripts and shell pipelines no longer treat these informational outputs as failures.
  • Added request timeouts to previously-unguarded requests calls in gget ref, gget info, gget 8cube, gget enrichr, and gget opentargets. Default is 10s connect / 60s read; configurable via the new DEFAULT_REQUESTS_TIMEOUT constant.
  • Narrowed a bare except: in utils.get_uniprot_seqs to (KeyError, IndexError, TypeError) so unrelated errors (including KeyboardInterrupt) are no longer swallowed.
  • Added utils.http_json() and utils.dig() helpers that issue a request and parse JSON / walk a nested response path with consistent error reporting. Migrated gget bgee, gget opentargets, and one .json() callsite in gget virus to use them; remaining modules will migrate opportunistically. Upstream HTML error pages, malformed JSON, and missing response keys now surface as clear RuntimeErrors naming the failing service instead of cryptic JSONDecodeError / KeyError tracebacks.
  • utils.http_json() now retries transient failures (connection errors, read timeouts, HTTP 5xx) up to 3 times with exponential backoff. Smooths over short upstream blips (e.g. bgee.org read timeouts) without affecting 4xx errors, which still raise immediately.
  • gget virus: Replaced 11 bare except: pass blocks around file.close() / os.remove() cleanup calls with narrowed except OSError handlers that log the failure at DEBUG. Previously, real I/O issues during cleanup (disk full, permissions) were silently dropped and the cleanup path also swallowed KeyboardInterrupt.
  • gget cbio: Fixed a code path in cbio_plot that called the removed-in-pandas-2.0 DataFrame.append() inside a loop when filling missing CNA genes — the entire branch crashed on modern pandas. It now builds a single DataFrame of missing rows and concatenates once.
• Performance:
  • utils.get_uniprot_seqs: Collect per-ID DataFrames in a list and pd.concat(..., ignore_index=True) once at the end, avoiding the O(n²) cost of growing a DataFrame inside the request loop.
  • Cached utils.find_latest_ens_rel, utils.search_species_options, utils.ref_species_options, and utils.find_nv_kingdom with functools.lru_cache. These hit Ensembl FTP listings that are stable for a release; repeated calls within one Python process are now free.
  • Added utils.parallel_map, a thin ThreadPoolExecutor wrapper for I/O-bound work. Used to fan out utils.get_uniprot_seqs across the input ID list — looking up N IDs is now bounded by ~N / pool_size UniProt round-trips instead of N. Pool size defaults to 8 and can be overridden via the GGET_MAX_WORKERS environment variable.