Same-GNSS guard in calc_pvt never throws (missing `throw`)

[zsoerenm]

Description

The guard in calc_pvt that is meant to reject mixed-constellation input never actually rejects anything — the ArgumentError is constructed but never thrown:

https://github.com/JuliaGNSS/PositionVelocityTime.jl/blob/master/src/PositionVelocityTime.jl#L238-L239

all(state -> state.system == states[1].system, states) ||
    ArgumentError("For now all satellites need to be base on the same GNSS")

Because condition || ArgumentError(...) only creates the exception object and discards it, a vector mixing GPS and Galileo satellites passes straight through. The code then proceeds with system = first(states).system and derives the week number / system start time from first(healthy_states).decoder only, silently producing a wrong result instead of erroring.

Fix

all(state -> state.system == states[1].system, states) ||
    throw(ArgumentError("For now all satellites need to be based on the same GNSS"))

(also fixes the "base" → "based" typo in the message).

Context

This is a pre-existing bug, surfaced while reviewing #38. That PR's constellation-wide ionospheric model selection assumes a single GNSS per solve, so enforcing this invariant becomes more relevant.