1.1:	a. Changed ALL references from lacv to lacf

	Primary reason: to prevent clashes between this package and costat

	Secondary reason: the autocovariance function in R for stationary
	series is called acf, so the time-varying one should be called
	lacf, should it not? 

	b. if the WPsmooth.type argument of lacf is "RM" then the
	behaviour of the binwidth argument is now different. If
	the argument binwidth is missing or set to zero then an automatic
	bandwidth is selected using the AutoBestBW function.

	c. Added a new argument to plot.lacf.

	The new argument is plot.it. If TRUE then a plot is produced,
	if FALSE then no plot is produced. In both cases the values
	that *would* have been plotted are returned (invisibly). In this
	way the plot.lacf function can be used as an extractor functon.

1.2:	Added plotcor=FALSE option for plot.lacf and plot.lacfCI for the
	type="line" plot. This permits plotting
	of the localized autocovariance function as a line plot

	Improved vertical (y) axis labelling by changing "ACF" to
	"Autocorrelation" and "ACF (cov)" to "Autocovariance"

	Changed lacf so that if lag.max=NULL then the same lag.max
	as in acf is used (this is 10*log(length(x)))

1.3:	changed code internally so that any xlab or ylab values supplied
	to plot.lacf or plot.lacfCI do not get obliterated.

1.4:	added references to paper where underpinning research was carried out

1.5:	added new function hwtos which performs the stationarity test for
	data which is of arbitrary length (not just power of two length
	data which is handled by hwtos2). The new hwtos function relies
	on the function hwt to compute the "arbitrary length" Haar wavelet
	transform which can be decimated or nondecimated. The hwtos
	function returns an object of class tosANYN and there are three
	new methods: plot.tosANYN, summary.tosANYN and print.tosANYN
	to plot, summarize and print the information

	The new function hwt computes the Haar wavelet transform on data
	with an arbitrary number of data points, not just a power of two.
	It can compute the decimated and nondecimated version. This
	transform can be plotted, printed and summarized with new methods.

1.6:	Added EstBetaCov function which computes the covariance matrix
	of the smoothed wavelet periodogram at a given location

	Added an extra message which catches the situation of a constant
	function more gracefully. The code is designed for stochastic
	series. [Thanks to a user of an earlier version for notifying
	me about this]

1.7:	corrected bug in calculation of covariance of raw
	wavelet periodogram. The covariance is only twice the square
	of the betas when ell=j and m=n

	This bug manifested itself mostly in plots of the confidence
	intervals for local autocorrelations. E.g. in Nason (2013)
	Figure 5, the confidence interval for the third lag coefficient
	was zero, and this is just wrong. After the bugfix these variances
	are estimated correctly and they are more plausible! 

	The defaults of AutoBestBW have been changed to tol=0.01 (from 0.1) and
	maxits=5 to maxits=200

1.7.1	Documented EstBetaCov function and added some argument checking.
	Corrected mismatch between documentation and code for AutoBestBW

	Removed hwtos function and its methods (not ready for prime time yet)
