Particles statistics fail with local time step

Particle statistics works fine with constant time step, but, in most cases, it leads to excessive calculation time. In lots of cases (cyclones, other separators, erosion) static and frozen flow approach is enough. In such a case using Steady (Local time step) option + frozen flow gives a huge performance advantage, but Saturne crashes instantly with memory violation error trying to calculate particle statistics in mesh cells. It's important that if you need gas-particles coupling (drying, combustion) your static case will not work at all with local timestep and particle statistics. This bug forces constant timestep for such conditions (flow + particles enabled) that rises calculation time incredibly. With frozen flow, workartound is a constant timestep (much slower but usable).
Steps to reproduce:
1. Use Saturne 8.0...8.3 versions (I think, any recent version).
2. Create or use ready case with gas and particles (no matter if the flow is frozen or not).
3. Set time stepping option to Steady + Local time step.
4. Enable particles statistics collection in particle tracking settings.
5. Run the case, Saturne will dump the core in the beginning of iteration.

使用启用粒子统计的 Steady + Local time step，通过 Saturne 8.0–8.3 的气体-粒子算例重现该问题。首先检查迭代开始时的核心转储，然后验证粒子统计不再触发内存违规，并验证该算例能够使用局部时间步运行。

由索引模型根据 Issue 内容生成。