Rename adix/stat -> mvstat; Fix nimble install

Big thing is stat -> mvstat rename; Right now the old name is still available but deprecated. Will probably be removed in 0.6.0.

Fix nimble install by moving cstats -> bu. Thanks to @Gruruya for pointing out the problem in #5 .

Updates, add tdigest, uniqce sketches

In more detail:

• Add merge API to lghisto
• Add tdigest tail quantile sketch.
• Add cardinality estimation sketch uniqce.nim
• Improve doc generation/re-do README to point to it.
• Rename adix/stat -> adix/mvstat

Even more in git log -p, of course.

B-tree & lfreq adjustments, slimSystem, add adix/amoft

Various B-tree fix-ups to make some things discardable/marked used/etc.

Allow user to boost util/lfreq data space, though it should be replaced by a less hokey thing when concepts work better..

Add new Count-Min Sketch-based adix/amoft.nim module for approximately k-most often problems, where right now k is not too large (probably need to tie into B-tree for large k and dynamically switch over for large k since the heap-scan-table is probably faster for k < 2000 which is likely the common case).

3 changes

Just:

• Get LPTabz.mmap compiling (but not quite working) under arc again. Works under refc, though.
• Track cligen findPathPattern for tests/anaPrime
• Add --uniq to util/lfreq

Fix bugs & add tests/wf-alike util/lfreq, util/cstats

Fix div by zero bugs, sequint.add from empty bug, in stat try to document why of SIMD-vectorized routines better and ease actually getting that boost, add stat.range for both min & max in 1-pass.

Track cligen pm=" +- " to just "+-" change in cstats and add it to the nimble install. Also add lfreq to the nimble install. Add cligen dependency for the two utilities.

Signature edits, warning clean-up, utility tweaks, add lfreq

The big patch is a change from z:static[int] to z:static int in signatures. Also a warning clean-up and a couple of tweaks to utilities tests/wf and util/cstats as well as adding util/lfreq which is what mostly motivates the timing of this release.

Cross Nim Portability, cstats, scope fix

Various cross Nim portability work (previewSlimSystem, BadIndex, tcc wf, etc.), Add ad hoc contextual statistics filter cstats, and a minor scope fix in adix/lptabz.nim.

Bug fixes & doc clean ups

Bug fixes are mostly Nim portability Re IndexDefect and lghisto quantiles.

Doc clean-ups are a lot & varied. git log has details.

Bug fixes & minor define-controls

Fix a bad bug of histogramming x instead of unshifted x0 and make
MovingStat work with OrderStats

Guard cligen usage with defined(useCligen) {to make .nimble not lie} and
also introduce -d:portablensort to deactivate Intel-specific PEXT32/64 nsort
optimizations. { Yes, yes..should just auto-detect/cpuid/etc. }

Make a couple test programs nicer.

`del` BREAKING CHANGE + new modules

There is a breaking change here. To be more drop-in semantically for the std/(tables|sets) world, lptabz.del & ditab.del no longer raise on missing keys. (Use missingOrExcl not exceptions if you want to test for that.) Sorry for the breaking change but cross/drop-in compatibility seems more important than backward compatibility at this stage of (guessed) user base. I must have not looked carefully at std/tables.del. Apologies.

Other changes:

• Speed-up adix/stat a bit. (backend C compilers need to at least ignore gcc __attribute__ ...).
• btree.add parameter order change to default queue discipline
• Rename bist -> Bist for general Nim convention consistency.
• New Bist-based module lghisto that is an alternative to things like tDigest but supports efficient tunable approximations to moving quantiles (where data changes point-by-point).
• New module cumsum for SIMD-optimized cumulative summation (sometimes called running sums or parallel prefix sum).
• New module nsort which does a optimized radix sort. I think the "bit level digit plan" using Intel's parallel bit extraction instruction may be a novel optimization or at least application of PEXT. (Byte level analogues probably date to early IBM sorting card research, though.) Bit-level should be great for things like float32 keys with data sets of similar order of magnitude keys probably being 2-pass. There are other standard optimizations in there, too. Read the module top.