FT_DEFINETRIAL defines the segments of data that will be used for
 further processing and analysis, i.e. the pieces of data that will
 be read in by FT_PREPROCESSING. Trials are defined by their begin
 and end sample in the data file and each trial has an offset that
 defines where the relative t=0 point (usually the sample at which
 the trigger is detected) is for that trial.

 Use as
   [cfg] = ft_definetrial(cfg)
 where the configuration structure should contain
   cfg.trialdef   = structure with details of trial definition, see below
   cfg.trialfun   = string with function name, see below (default = 'ft_trialfun_general')
 and furthermore
   cfg.dataset     = string with the filename
   cfg.headerfile  = string with the filename
   cfg.datafile    = string with the filename
 and optionally

 A call to FT_DEFINETRIAL results in the trial definition "trl" being
 added to the output configuration structure. The trials are defined
 according to the triggers, trials or other events in the data, or
 from a user-specified MATLAB function (subsequently referred to as
 the trial function) which returns "trl". The user can specify the
 name of his/her custom trial function that is tailored to the
 experimental paradigm, or use the default trial function

 The trial definition "trl" is an Nx3 matrix, N is the number of trials.
 The first column contains the sample-indices of the begin of each trial
 relative to the begin of the raw data, the second column contains the
 sample-indices of the end of each trial, and the third column contains
 the offset of the trigger with respect to the trial. An offset of 0
 means that the first sample of the trial corresponds to the trigger. A
 positive offset indicates that the first sample is later than the trigger,
 a negative offset indicates that the trial begins before the trigger.

 The trial definition "trl" can contain additional columns besides the
 required three that represent begin, end and offset. These additional
 columns can be used by a custom trialfun to provide numeric information
 about each trial such as trigger codes, response latencies, trial
 type and response correctness. The additional columns of the "trl"
 matrix will be represented in data.trialinfo after FT_PREPROCESSING.

 If FT_TRIALFUN_GENERAL has been used to generate the "trl" matrix, the
 function may return a fourth column in the trl-matrix, that refers to
 the event-code for the corresponding trial. It depends on the acquisition
 system whether or not this fourth column is returned. In general, this
 fourth column is generated by default if the event codes are represented
 numerically, or as a string starting with 'S' or 'R' (for BrainVision data)

 Simple trial definitions (e.g. based on a single trigger) are supported
 by FT_TRIALFUN_GENERAL, which is the default trial function. This function
 supports the following options
   cfg.trialdef.eventtype  = 'string'
   cfg.trialdef.eventvalue = number, string or list with numbers or strings
   cfg.trialdef.prestim    = number, latency in seconds (optional)
   cfg.trialdef.poststim   = number, latency in seconds (optional)

 If you specify cfg.trialdef.eventtype  = '?' a list with the events in your
 data file will be displayed on screen.

 If you want to read all data from a continuous file in a single or in
 multiple segments, TRIALFUN_GENERAL understands the following options
    cfg.trialdef.triallength = duration in seconds (can also be 1 or Inf)
    cfg.trialdef.ntrials     = number of trials (can also be 1 or Inf)

 If you need to define the segments of interest on the basis of a conditional
 sequence of events (e.g. stimulus trigger followed by a correct response),
 you should supply your own trial function and specify that as cfg.trialfun.
 See below for pointers to some examples.

 Both the default trial function and your own supplied custom trial
 functions in general will call FT_READ_EVENT to collect all event
 information (such as triggers) from your dataset and to select
 pieces of data according to this information.

 The cfg.trialfun option is a string containing the name of a function
 that you wrote yourself and that FT_DEFINETRIAL will call. The
 function should take the cfg-structure as input and should give a
 NxM matrix with M equal to or larger than 3) in the same format as
 "trl" as the output. You can add extra custom fields to the
 configuration structure to pass as arguments to your own trialfun.
 Furthermore, inside the trialfun you can use the FT_READ_EVENT
 function to get the event information from your data file.