Fixes parameter issues raised in #600 (except CombinedParameter), and maybe also #498, #92. See below for list of open questions

Major parameter changes

• Remove ManualParameter (it's behaviour can be captured quite easily in Parameter). Keep as deprecated parameter.
• Remove StandardParameter,, as most of it properties should be used by a wider range of parameters. Its unique feature is connecting a function/VISA command to its get/set operations. This can be incorporated into the Parameter. Keep as deprecated parameter
• Create wrappers for get/set that ensures additionals such as delays are always included

Other changes

• Change behaviour of parameter.get_latest(). It now will return the latest value, unless it was obtained longer than max_val_age ago, in which case it will perform a new get(). This ensures that a value is never used that is older than max_val_age. Part of this includes moving max_val_age into GetLatest.
• Remove max_delay, delay_tolerance as they were not doing anything. If anyone knows what they are supposed to do please let me know
• Modify get/set_delay into property
• Change delay into post_delay, i.e. delay (settling time) after each get()
• Add inter_delay, i.e. minimum delay between successive get() calls
• Move delays to BaseParameter, as they can be useful for other parameter types as well
• Change get_step/set_step to property
• Move step to BaseParameter, as it could also be useful for other parameters (such as CombinedParameter or ManualParameter)
• Change _set_get/_set_set to _initialize_get/_initialize_set because naming was confusing
• Combine validate_and_set, validate_and_sweep to set()
• Add get_sweep_values method to BaseParameter, used when performing a .set. It's default behaviour is to simply return the same value, but if step is defined, it will return a list of sweep values (taken from StandardParameter).
• Move validate to BaseParameter, as it later might be used by other parameter types as well.
• Accept numpy arrays as setpoints (Fixes one of the MultiParameter issues in MultiParameter usability issues #498, other issues need to be fixed elsewhere)
• Add scale, which can scale (convert) its value. This can be handled in the get/set wrappers. Additionally, a raw_value should be an attribute which is included in the metadata
• Accept numpy array for sweep points. Not sure if this should be via slicing (might be removed SweepValues, Troubles ?  #465), or .sweep, although this might be too long (param.sweep(vals=sweep_vals) vs param[sweep_vals]. Required changing SweepFixedValues
• Update tests, include new tests.
• Removal of units in (Array)Parameter, as it was already deprecated
• Remove full_name, replace by str(parameter). full_name is kept with deprecation warning
• Combine _latest_val, _latest_ts into _latest dict attr, remove _latest method
• Remove no_getter/setter, as they are no longer used
• Only attributes in meta_attrs that are not None are included. As there are now more possible attrs that are by default included, this could otherwise lead to significant overhead when obtaining a snapshot of all parameters.

Open questions

• Should parameters include logging through the logging module? I'm asking because it might lead to excessive logging (in measurements, parameters can be set hundreds of thousands of times).
• Should signalling be added to notify parameter event changes (Use events to couple parameters in instrument drivers? #92)? This could have several useful features. One example is that instruments can connect to a parameter, such that if the parameter changes its value, the instrument can raise a flag that it needs to be setup (New "Configuration Parameter"  #377). Using a package like blinker works well for this purpose, but only in a single process (i.e. it does not by itself support multiprocessing). However, some sort of message passing can probably be used to incorporate this. @giulioungaretti do you think this is worthwhile? It can be easily included in the get/set wrapper
• How can I test for things like delays? Won't including delays in the TestSuite slow down testing?

@giulioungaretti @jenshnielsen @WilliamHPNielsen @MerlinSmiles