Reverted buggy commit 5366dd5cb9f546809ff21a0bd23a2ac423aac6ce by Ferhat and added call to strtrim in @dynSeries/extract method to remove leading and trailing white spaces in variable names.

Modified @dynSeries/mrdivide method so that dynSeries objects defined over different non disjoint periods can be divided.

Modified @dynSeries/mtimes method so that dynSeries objects defined over different non disjoint periods can be multiplied.

Modified @dynSeries/minus method so that dynSeries objects defined over different non disjoint periods can be substracted.

Modified @dynSeries/plus method so that dynSeries objects defined over different non disjoint periods can be added.

As in dynSeries/extract we consider a tensor product.

*Example*

Suppose that a dynSeries object contains the following set of variables

GDP_1
GDP_2
GDP_3
GDP_4
GDP_5
GDP_6
GDP_7
HICP_1
HICP_2
HICP_3
HICP_4
HICP_5
HICP_6
HICP_7
Consumption_1
Consumption_2
Consumption_3
Consumption_4
Consumption_5
Consumption_6
Consumption_7

Then the syntax:

b = a{'@GDP,HICP@_@1,2,3@'}

will instantiate a new dynSeries object containing variables:

GDP_1
GDP_2
GDP_3
HICP_1
HICP_2
HICP_3

Added the possibility to extract variables from dynSeries object using indices instead of variable names.

If ts is a dynSeries object containing variables A1, A2, A3, A4, A5, then the following syntaxes are equivalent:

ts{[1, 2, 5]}

ts{'A1','A2','A5'}

and will create a new dynSeries object made of variables A1, A2 and A5.

If called with a second input argument (dim), size returns the number of observations if dim==1 or
the number of variables if dim==2. Without this second argument, size returns the number of
observations and the number of variables.

In addition to the already implemented selection mechanism using matlab's regular expressions,
we allow the following syntax:

a = ts{'GDP_@US,JP,FR@'};

to extract variables GDP_US, GDP_JP and GDP_FR from dynSeries object ts. If one of these variables
does not exist, an error message is issued.

Names for the results of basic operations on dynSeries objects was '--NA--'. The name of the result
variables now keep track of the operations. For instance, if we have two dynSeries objects, ts1 and
ts2, respectively holding two variables 'A1' and 'B1', then:

 - the name of the variable in ts3=ts1+ts2 is 'plus(A1,B1)'
 - the name of the variable in ts3=ts1-ts2 is 'minus(A1,B1)'
 - the name of the variable in ts3=ts1*ts2 is 'multiply(A1,B1)'
 - the name of the variable in ts3=ts1/ts2 is 'divide(A1,B1)'
 - the tex name of the variable in ts3=ts1+ts2 is '(A1+B1)'
 - the tex name of the variable in ts3=ts1-ts2 is '(A1-B1)'
 - the tex name of the variable in ts3=ts1*ts2 is '(A1*B1)'
 - the tex name of the variable in ts3=ts1/ts2 is '(A1/B1)'

Returns zero if the input argument is a well defined dynSeries object.