Clean tspa io and some fixes

[ldeniau]

The particle type detection based on relative error is extended to 1e-4.
Clean TPSA I/O (PR restored), backward compatible (default madxprint=false in a_scratch_size.f90, used only by TPSA low level I/O routines in [c]c_dabnew.f90).
mapdump add one more level to dump maps to all orders (was limited to 1st order before).

[ldeniau]

A comment about maxacceleration vs cavsareset in madx_ptc_module.f90. By default the code set cavreset to false independently of maxacceleration, while in some studies (MAD-NG energy patches for Linac4) I had to setup the value in a different way to get it working in MAD-X. The code is there but commented (for future review).

[rdemaria]

Ok for me. Maybe Piotr could give a look...

[ldeniau]

Waiting for Piotr comment

[piotrskowronski]

This functionality was implemented for CTF3, RIP. I don't know any other machine that uses it.
It was never bulletproof and I know that at certain PTC version it become problematic. The issues are related to handling time true/false and totalpath true/false that this routine switches back and forth.

Therefore, I believe that it is wise to remove it and eventually re-design and re-implement it from scratch if some user requires it.

Please forgive me my delayed reaction, but since I left the group I keep ignoring the notifications from github.