nwchem.scalar: fix S2 regex matching any line containing digit 2

The active S2 branch was "elsif ( /S*2/ )" (S* matches empty, so this
is really just "line contains a 2 anywhere") -- not the commented-out
"/S\s*2/" above it, and not what real NWChem output actually prints
either (the real tag is literally "S^2", confirmed against a real
UHF ecce_print trace). Worked "by luck" only because no other property
routed through this shared script (EGRAD/CPUSEC/ETIME/NNREPUL/DELTAE/
RESIDNORM/ONEELEC/TWOELEC/CORRELTN/COULOMB/EXCHANGE) has a literal "2"
in its header tag text -- S2 extraction itself was never actually
gated on the S2 marker matching.

Fixed to /S\^2/, matching the evident original intent. Verified: real
NWChem 7.2.3 UHF OH/STO-3G run extracts S2=0.753263362179062, matching
Gaussian's S2=0.753262 for the same system/basis (independent cross-
code sanity check).

Also documents NWChem's ecce_print mechanism in CLAUDE.md -- found
while auditing nwchem.desc, whose Begin patterns look meaningless
against raw stdout because they're not matched against stdout at all;
they match a separate trace file NWChem writes via its own built-in
ecce_print directive. Non-obvious enough to mislead a future session
auditing this file without discovering it first (this session initially
didn't either).

Part of a broader Gaussian-16/NWChem parser audit (real g16/nwchem
jobs, eccejobmonitor's actual Begin/Skip/Lines/End/Frequency algorithm
simulated against real captured output) that found no other live bugs
-- several confirmed-dead-but-harmless .desc entries documented in the
issue tracker/session notes rather than "fixed" (e.g. gaussian-16.desc's
EMP21/EMP22 Begin collision produces identical output either way;
gaussian-16.desc's standalone CPUSEC entry never fires but the same
data is already extracted via a different code path, gaussian-16.db).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Author: andy

CLAUDE.md

@@ -74,6 +74,15 @@ accurate) for the full pipeline; short version:
   close to what you need rather than inventing a new shape).
 - **`scripts/gensub`** / **`scripts/eccejobmonitor`** — job-submission
   script generation and remote progress reporting.
+- **NWChem's `.desc` `Begin` patterns (`%begin%`/`%end%`/`task_*`) don't
+  match raw NWChem stdout at all, and look bizarre if you try** — they
+  match a *separate*, machine-tagged trace file that NWChem writes via
+  its own built-in `ecce_print <file>` directive (still supported as of
+  NWChem 7.2.3, confirmed live), wired up by `nwch.tpl`'s `ecce_print
+  ##parseFile##` line and `nwchem.launchpp` (rewrites that line to an
+  absolute path at launch time). This *is* the file `eccejobmonitor`
+  actually reads for NWChem — auditing `nwchem.desc` against plain
+  stdout instead will wrongly conclude the whole file is dead.
 - **The one gotcha that has bitten this project repeatedly**: every file
   under `scripts/*` needs its own `install()` rule in `CMakeLists.txt`,
   or it's simply absent from the packaged `.deb` — correct in the repo,

scripts/parsers/nwchem.scalar

@@ -61,7 +61,7 @@ if ( /begin\%gradient norm/ ) {
   $unit = "units:\nHartree\n";
 # } elsif ( /S\s*2/ ) {
 #  $propKey = "S2";
-} elsif ( /S*2/ ) {
+} elsif ( /S\^2/ ) {
   $propKey = "S2";
 }
 $value = <STDIN>;