fix(gpscnav): don't require T_GD for positioning readiness

[giove-a]

Summary

is_clock_correction_decoded(::GPSCNAVData) gated positioning readiness on
T_GD. Since T_GD is broadcast only in CNAV message type 30 — separately
from, and far less often than, the SV clock — a receiver could fully decode a
satellite's ephemeris and clock yet never report it usable for positioning. This
drops the T_GD requirement (the SV clock polynomial is sufficient for a fix);
T_GD is still applied downstream when present. Affects GPS L2C and L5I
(shared CNAV core).

Root cause

is_decoding_completed_for_positioning → is_clock_correction_decoded required
t_0c, a_f0, a_f1, a_f2 and T_GD. The clock polynomial rides in the shared
clock block of any of MT30–37 (parse_clock_block), but T_GD is set only by
parse_mt30. Positioning readiness was therefore silently coupled to MT30 being
in the broadcast schedule.

On a CNAV stream that carries the clock via MT31–37 without MT30, every message
decodes and the ephemeris + clock complete, but positioning never validates — and
because validate_data is gated on the same check, raw_data is never promoted
to data (and the streaming counter never re-armed), so nothing downstream works
either.

Why T_GD must not gate a fix (ICD)

T_GD is a ~metre-level inter-signal (group-delay) correction, not required for a
position solution, and it is structurally infrequent. Maximum broadcast intervals:

data	message type(s)	L2C (IS-GPS-200 Tbl 30-XII)	L5 (IS-GPS-705J Tbl 20-XII)
ephemeris	10 & 11	48 s	24 s
clock	any of 30–37	48 s	24 s
T_GD (in ISC/IONO)	30 only	288 s	144 s

So T_GD is guaranteed only ~6× less often than the clock/ephemeris — requiring
it can stall a fix for up to 288 s (L2C) / 144 s (L5) on a fully spec-compliant
satellite.

For reference within the codebase: GPS L1C-D (CNAV-2) already does not gate
positioning on T_GD/ISC (only TOI + subframe-2 CED), and LNAV is unaffected
(T_GD rides in subframe 1 alongside the clock). This change brings CNAV in line.

The fix

Remove !isnothing(data.T_GD) from is_clock_correction_decoded; require only
the SV clock polynomial (t_0c, a_f0, a_f1, a_f2). One function, plus a comment
documenting the ICD rationale, and a note on the GPSCNAVData docstring that the
MT30-only fields (T_GD, ISCs, Klobuchar α/β) may be nothing even once
positioning is ready.

Validation (real + simulated captures)

• TEXCUP (UT Austin, real GPS, NTLAB front-end): before — L2C stalls on
  satellites whose MT30 lags; after — L2C produces a continuous PVT that agrees
  with the independent L1 C/A solution to < 1 m (and the L2C-decoded orbit
  matches the L1 broadcast to 0.82 m).
• Fraunhofer Flexiband (Spirent simulation whose CNAV schedule omits MT30):
  before — all 10 L2C and all 10 L5I satellites decode a full ephemeris but 0
  position (T_GD never arrives); after — they validate and position, all three
  bands (L1 / L2C / L5I) landing on the Fraunhofer IIS site to ~0.2 m.

Testing

• Pkg.test() — 2446/2446 pass, including Aqua.test_all.
• JuliaFormatter 2.8.5 (repo config) — clean, no diff.
• Existing CNAV tests unaffected: full-decode / decode_once fixtures carry MT30,
  and the !is_decoding_completed_for_positioning assertion in the L5I tests is
  gated by the ephemeris check (MT11 absent), not the clock check.

Related (not in this PR)

On MT30-less streams PositionVelocityTime still needs the MT30-only corrections
T_GD, ISC_L2C / ISC_L5I5, and the Klobuchar α/β, and throws
+(::Float64, ::Nothing) when they're absent. A companion PVT change to treat
these as absent-→-zero (and skip iono when α/β are missing) would let L2C/L5
position end-to-end without a workaround.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.15%. Comparing base (7610c18) to head (98c5050).

Additional details and impacted files

@@           Coverage Diff           @@
##           master      #68   +/-   ##
=======================================
  Coverage   98.15%   98.15%           
=======================================
  Files          14       14           
  Lines        1515     1515           
=======================================
  Hits         1487     1487           
  Misses         28       28           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.