Don't cache failed queries

Some generic solution were added to `query.py` in https://github.com/astropy/astroquery/pull/2187, but all the modules has to be checked and updated if needed to use it.

- [ ] alfalfa
- [ ] alma
- [ ] astrometry_net
- [ ] atomic
- [ ] besancon
- [ ] cadc
- [ ] casda
- [ ] cds
- [ ] cosmosim
- [ ] dace
- [ ] esa/hubble
- [ ] esa/iso
- [ ] esa/xmm_newton
- [ ] esasky
- [ ] eso
- [ ] exoplanet_orbit_database
- [ ] fermi
- [ ] gaia
- [ ] gama
- [ ] gemini
- [ ] heasarc
- [ ] hips2fits
- [ ] hitran
- [ ] image_cutouts/first
- [ ] ipac/irsa
- [ ] ipac/irsa/ibe
- [ ] ipac/irsa/irsa_dust
- [ ] ipac/irsa/sha
- [ ] ipac/ned
- [ ] ipac/nexsci/nasa_exoplanet_archive
- [ ] jplspec
- [ ] lamda
- [ ] magpis
- [ ] mast
- [ ] mpc
- [ ] nasa_ads
- [ ] nasa_exoplanet_archive
- [ ] ned
- [ ] nist
- [ ] noirlab
- [ ] nrao
- [ ] nvas
- [ ] oac
- [ ] ogle
- [ ] sdss
- [ ] simbad
- [ ] skyview
- [ ] solarsystem/imcce
- [ ] solarsystem/imcce/miriade
- [ ] solarsystem/imcce/skybot
- [ ] solarsystem/jpl/horizons
- [ ] solarsystem/jpl/sbdb
- [ ] solarsystem/mpc
- [ ] splatalogue
- [ ] svo_fps
- [ ] template_module
- [ ] ukidss
- [ ] vamdc
- [ ] vizier
- [ ] vo_conesearch
- [ ] vsa
- [ ] wfau
- [ ] xmatch

Start with query.py and pull request #2187 to understand the generic failed-query handling. Then review the modules listed in the checklist and identify which ones need to use it; done means every applicable module has been checked and updated as needed.

Written by the indexing model from the issue text.